WorkflowGeneratorSteps.cs

using System;
using FreneticUtilities.FreneticExtensions;
using Newtonsoft.Json.Linq;
using SwarmUI.Core;
using SwarmUI.Media;
using SwarmUI.Text2Image;
using SwarmUI.Utils;

using static SwarmUI.Builtin_ComfyUIBackend.WorkflowGenerator;

namespace SwarmUI.Builtin_ComfyUIBackend;

public class WorkflowGeneratorSteps
{
    /// <summary>Register a new step to the workflow generator.</summary>
    public static void AddStep(Action<WorkflowGenerator> step, double priority)
    {
        WorkflowGenerator.AddStep(step, priority);
    }

    /// <summary>Register a new step to the workflow generator.</summary>
    public static void AddModelGenStep(Action<WorkflowGenerator> step, double priority)
    {
        WorkflowGenerator.AddModelGenStep(step, priority);
    }

    /* ========= RESERVED NODES ID MAP =========
     * 4: Initial Model Loader
     * 5: VAE Encode Init or Empty Latent
     * 6: Positive Prompt
     * 7: Negative Prompt
     * 8: Final VAEDecode
     * 9: Final Image Save
     * 10: Main KSampler
     * 11: Alternate Main VAE Loader
     * 15: Image Load
     * 20: Refiner Model Loader
     * 21: Refiner VAE Loader
     * 23: Refiner KSampler
     * 24: Refiner VAEDecoder
     * 25: Refiner VAEEncode
     * 26: Refiner ImageScale
     * 27: Refiner UpscaleModelLoader
     * 28: Refiner ImageUpscaleWithModel
     * 29: Refiner ImageSave
     * 30: Video Image Pre-Save
     *
     * 100+: Dynamic
     * 1500+: LoRA Loaders (Stable-Dynamic)
     * 50,000+: Intermediate Image Saves (Stable-Dynamic)
     */

    public static void Register()
    {
        #region Model Loader
        AddStep(g =>
        {
            g.FinalLoadedModel = g.UserInput.Get(T2IParamTypes.Model);
            if (g.FinalLoadedModel is null)
            {
                throw new SwarmUserErrorException($"No model input given. Did your UI load properly?");
            }
            g.FinalLoadedModelList = [g.FinalLoadedModel];
            (g.FinalLoadedModel, g.CurrentModel, g.CurrentTextEnc, g.CurrentVae) = g.CreateModelLoader(g.FinalLoadedModel, "Base", "4", sectionId: T2IParamInput.SectionID_BaseOnly);
        }, -15);
        AddModelGenStep(g =>
        {
            if (g.IsRefinerStage && g.UserInput.TryGet(T2IParamTypes.RefinerVAE, out T2IModel rvae))
            {
                g.LoadingVAE = g.CreateVAELoader(rvae.ToString(g.ModelFolderFormat), g.HasNode("21") ? null : "21");
            }
            else if (!g.NoVAEOverride && g.UserInput.TryGet(T2IParamTypes.VAE, out T2IModel vae))
            {
                if (g.FinalLoadedModel.ModelClass?.ID == "stable-diffusion-v3-medium" && vae.ModelClass?.CompatClass?.ID != "stable-diffusion-v3")
                {
                    Logs.Warning($"Model {g.FinalLoadedModel.Title} is an SD3 model, but you have VAE {vae.Title} selected. If that VAE is not an SD3 specific VAE, this is likely a mistake. Errors may follow. If this breaks, disable the custom VAE.");
                }
                g.LoadingVAE = g.CreateVAELoader(vae.ToString(g.ModelFolderFormat), g.HasNode("11") ? null : "11");
            }
            else if (!g.NoVAEOverride && g.UserInput.Get(T2IParamTypes.AutomaticVAE, false))
            {
                string clazz = g.FinalLoadedModel.ModelClass?.CompatClass?.ID;
                string vaeName = null;
                if (clazz == "stable-diffusion-xl-v1")
                {
                    vaeName = g.UserInput.SourceSession?.User?.Settings.VAEs.DefaultSDXLVAE;
                }
                else if (clazz == "stable-diffusion-v1")
                {
                    vaeName = g.UserInput.SourceSession?.User?.Settings.VAEs.DefaultSDv1VAE;
                }
                if (!string.IsNullOrWhiteSpace(vaeName) && vaeName.ToLowerFast() != "none")
                {
                    string match = T2IParamTypes.GetBestModelInList(vaeName, Program.T2IModelSets["VAE"].ListModelNamesFor(g.UserInput.SourceSession));
                    if (match is not null)
                    {
                        T2IModel vaeModel = Program.T2IModelSets["VAE"].Models[match];
                        g.LoadingVAE = g.CreateVAELoader(vaeModel.ToString(g.ModelFolderFormat), g.HasNode("11") ? null : "11");
                    }
                }
            }
        }, -14);
        AddModelGenStep(g =>
        {
            (g.LoadingModel, g.LoadingClip) = g.LoadLorasForConfinement(-1, g.LoadingModel, g.LoadingClip);
            (g.LoadingModel, g.LoadingClip) = g.LoadLorasForConfinement(0, g.LoadingModel, g.LoadingClip);
            if (g.IsRefinerStage)
            {
                (g.LoadingModel, g.LoadingClip) = g.LoadLorasForConfinement(1, g.LoadingModel, g.LoadingClip);
            }
            else if (g.IsImageToVideoSwap)
            {
                (g.LoadingModel, g.LoadingClip) = g.LoadLorasForConfinement(T2IParamInput.SectionID_VideoSwap, g.LoadingModel, g.LoadingClip);
            }
            else if (g.IsImageToVideo)
            {
                (g.LoadingModel, g.LoadingClip) = g.LoadLorasForConfinement(T2IParamInput.SectionID_Video, g.LoadingModel, g.LoadingClip);
            }
            else // Base
            {
                (g.LoadingModel, g.LoadingClip) = g.LoadLorasForConfinement(T2IParamInput.SectionID_BaseOnly, g.LoadingModel, g.LoadingClip);
            }
        }, -10);
        AddModelGenStep(g =>
        {
            if (g.UserInput.TryGet(ComfyUIBackendExtension.SetClipDevice, out string device) && g.Features.Contains("set_clip_device"))
            {
                string clipDeviceNode = g.CreateNode("OverrideCLIPDevice", new JObject()
                {
                    ["clip"] = g.LoadingClip,
                    ["device"] = device
                });
                g.LoadingClip = [clipDeviceNode, 0];
            }
        }, -9);
        AddModelGenStep(g =>
        {
            string applyTo = g.UserInput.Get(T2IParamTypes.FreeUApplyTo, null);
            if (g.Features.Contains("freeu") && applyTo is not null)
            {
                if (applyTo == "Both" || applyTo == g.LoadingModelType)
                {
                    string version = g.UserInput.Get(T2IParamTypes.FreeUVersion, "1");
                    string freeU = g.CreateNode(version == "2" ? "FreeU_V2" : "FreeU", new JObject()
                    {
                        ["model"] = g.LoadingModel,
                        ["b1"] = g.UserInput.Get(T2IParamTypes.FreeUBlock1),
                        ["b2"] = g.UserInput.Get(T2IParamTypes.FreeUBlock2),
                        ["s1"] = g.UserInput.Get(T2IParamTypes.FreeUSkip1),
                        ["s2"] = g.UserInput.Get(T2IParamTypes.FreeUSkip2)
                    });
                    g.LoadingModel = [freeU, 0];
                }
            }
        }, -8);
        AddModelGenStep(g =>
        {
            if (g.UserInput.TryGet(ComfyUIBackendExtension.SelfAttentionGuidanceScale, out double sagScale))
            {
                string patched = g.CreateNode("SelfAttentionGuidance", new JObject()
                {
                    ["model"] = g.LoadingModel,
                    ["scale"] = sagScale,
                    ["blur_sigma"] = g.UserInput.Get(ComfyUIBackendExtension.SelfAttentionGuidanceSigmaBlur, 2.0)
                });
                g.LoadingModel = [patched, 0];
            }
            if (g.UserInput.TryGet(ComfyUIBackendExtension.PerturbedAttentionGuidanceScale, out double pagScale))
            {
                string patched = g.CreateNode("PerturbedAttentionGuidance", new JObject()
                {
                    ["model"] = g.LoadingModel,
                    ["scale"] = pagScale
                });
                g.LoadingModel = [patched, 0];
            }
            if (g.UserInput.TryGet(ComfyUIBackendExtension.RescaleCFGMultiplier, out double rescaleCfg))
            {
                string patched = g.CreateNode("RescaleCFG", new JObject()
                {
                    ["model"] = g.LoadingModel,
                    ["multiplier"] = rescaleCfg
                });
                g.LoadingModel = [patched, 0];
            }
            if (g.UserInput.TryGet(ComfyUIBackendExtension.RenormCFG, out double renormCfg))
            {
                string patched = g.CreateNode("RenormCFG", new JObject()
                {
                    ["model"] = g.LoadingModel,
                    ["cfg_trunc"] = 100, // This value is the weirdly named timestep where the renorm applies - less than this apply, above don't. 100 is default, not sure if it needs a customization param?
                    ["renorm_cfg"] = renormCfg
                });
                g.LoadingModel = [patched, 0];
            }
            if (g.UserInput.Get(ComfyUIBackendExtension.UseCfgZeroStar, false))
            {
                string patched = g.CreateNode("CFGZeroStar", new JObject()
                {
                    ["model"] = g.LoadingModel
                });
                g.LoadingModel = [patched, 0];
            }
            if (g.UserInput.Get(ComfyUIBackendExtension.UseTCFG, false))
            {
                string patched = g.CreateNode("TCFG", new JObject()
                {
                    ["model"] = g.LoadingModel
                });
                g.LoadingModel = [patched, 0];
            }
            if (g.UserInput.TryGet(ComfyUIBackendExtension.NormalizedAttentionGuidanceScale, out double nagScale) && nagScale > 0)
            {
                string patched = g.CreateNode("NAGuidance", new JObject()
                {
                    ["model"] = g.LoadingModel,
                    ["nag_scale"] = nagScale,
                    ["nag_alpha"] = g.UserInput.Get(ComfyUIBackendExtension.NormalizedAttentionGuidanceAlpha, 0.5),
                    ["nag_tau"] = g.UserInput.Get(ComfyUIBackendExtension.NormalizedAttentionGuidanceTau, 1.5),
                });
                g.LoadingModel = [patched, 0];
            }
        }, -7);
        AddModelGenStep(g =>
        {
            if (g.UserInput.TryGet(T2IParamTypes.ClipStopAtLayer, out int layer))
            {
                string clipSkip = g.CreateNode("CLIPSetLastLayer", new JObject()
                {
                    ["clip"] = g.LoadingClip,
                    ["stop_at_clip_layer"] = layer
                });
                g.LoadingClip = [clipSkip, 0];
            }
        }, -6);
        AddModelGenStep(g =>
        {
            if (g.UserInput.TryGet(T2IParamTypes.SeamlessTileable, out string tileable) && tileable != "false")
            {
                string mode = "Both";
                if (tileable == "X-Only") { mode = "X"; }
                else if (tileable == "Y-Only") { mode = "Y"; }
                string tiling = g.CreateNode("SwarmModelTiling", new JObject()
                {
                    ["model"] = g.LoadingModel,
                    ["tile_axis"] = mode
                });
                g.LoadingModel = [tiling, 0];
                string tilingVae = g.CreateNode("SwarmTileableVAE", new JObject()
                {
                    ["vae"] = g.LoadingVAE,
                    ["tile_axis"] = mode
                });
                g.LoadingVAE = [tilingVae, 0];
            }
        }, -5);
        AddModelGenStep(g =>
        {
            if (g.UserInput.TryGet(ComfyUIBackendExtension.TeaCacheMode, out string teaCacheMode) && teaCacheMode != "disabled")
            {
                double teaCacheThreshold = g.UserInput.Get(ComfyUIBackendExtension.TeaCacheThreshold, 0.25);
                double teaCacheStart = g.UserInput.Get(ComfyUIBackendExtension.TeaCacheStart, 0);
                if (teaCacheMode == "base gen only" && g.LoadingModelType != "Base")
                {
                    // wrong step, skip
                }
                else if (g.FinalLoadedModel?.Metadata?.SpecialFormat == "nunchaku" || g.FinalLoadedModel?.Metadata?.SpecialFormat == "nunchaku-fp4")
                {
                    Logs.Warning($"Ignore TeaCache Mode parameter because the current model is Nunchaku which does not support TeaCache. Use 'Nunchaku Cache Threshold' for a similar effect to TeaCache.");
                }
                else if (g.IsFlux())
                {
                    if (teaCacheMode != "video only")
                    {
                        string teaCacheNode = g.CreateNode(g.Features.Contains("teacache_oldvers") ? "TeaCacheForImgGen" : "TeaCache", new JObject()
                        {
                            ["model"] = g.LoadingModel,
                            ["model_type"] = "flux",
                            ["rel_l1_thresh"] = teaCacheThreshold,
                            ["max_skip_steps"] = 3,
                            ["start_percent"] = teaCacheStart,
                            ["end_percent"] = 1,
                            ["cache_device"] = "cuda"
                        });
                        g.LoadingModel = [teaCacheNode, 0];
                    }
                }
                else if (g.IsHunyuanVideo() || g.IsLTXV() || g.IsWanVideo() || g.IsHiDream())
                {
                    string type = "";
                    if (g.IsHunyuanVideo())
                    {
                        type = "hunyuan_video";
                    }
                    else if (g.IsLTXV())
                    {
                        type = "ltxv";
                    }
                    else if (g.IsHiDream())
                    {
                        type = "hidream_i1_dev";
                    }
                    else
                    {
                        string arch = g.CurrentModelClass()?.ID;
                        if (arch == "wan-2_1-text2video-1_3b" || arch == "wan-2_1-image2video-1_3b")
                        {
                            type = "wan2.1_t2v_1.3B";
                        }
                        else if (arch == "wan-2_1-text2video-14b")
                        {
                            type = "wan2.1_t2v_14B";
                        }
                        else if (arch == "wan-2_1-image2video-14b" || arch == "wan-2_1-flf2v-14b")
                        {
                            if (g.FinalLoadedModel.Name.Contains("720p") || g.FinalLoadedModel.StandardWidth == 960 || arch == "wan-2_1-flf2v-14b")
                            {
                                type = "wan2.1_i2v_720p_14B";
                            }
                            else
                            {
                                type = "wan2.1_i2v_480p_14B";
                            }
                        }
                    }
                    string teaCacheNode = g.CreateNode(g.Features.Contains("teacache_oldvers") ? "TeaCacheForVidGen" : "TeaCache", new JObject()
                    {
                        ["model"] = g.LoadingModel,
                        ["model_type"] = type,
                        ["rel_l1_thresh"] = teaCacheThreshold,
                        ["max_skip_steps"] = 3,
                        ["start_percent"] = teaCacheStart,
                        ["end_percent"] = 1,
                        ["cache_device"] = "cuda"
                    });
                    g.LoadingModel = [teaCacheNode, 0];
                }
                else
                {
                    Logs.Debug($"Ignore TeaCache Mode parameter because the current model is '{g.CurrentModelClass()?.Name ?? "(none)"}' which does not support TeaCache.");
                }
            }
            if (g.UserInput.TryGet(ComfyUIBackendExtension.EasyCacheMode, out string easyCacheMode) && easyCacheMode != "disabled")
            {
                if (teaCacheMode == "base gen only" && g.LoadingModelType != "Base")
                {
                    // wrong step, skip
                }
                else if (g.IsVideoModel() || teaCacheMode != "video only")
                {
                    string teaCacheNode = g.CreateNode("EasyCache", new JObject()
                    {
                        ["model"] = g.LoadingModel,
                        ["reuse_threshold"] = g.UserInput.Get(ComfyUIBackendExtension.EasyCacheThreshold, 0),
                        ["start_percent"] = g.UserInput.Get(ComfyUIBackendExtension.EasyCacheStart, 0),
                        ["end_percent"] = g.UserInput.Get(ComfyUIBackendExtension.EasyCacheEnd, 1),
                        ["verbose"] = false
                    });
                    g.LoadingModel = [teaCacheNode, 0];
                }
            }
        }, -4);
        AddModelGenStep(g =>
        {
            if (g.Features.Contains("aitemplate") && g.UserInput.Get(ComfyUIBackendExtension.AITemplateParam))
            {
                string aitLoad = g.CreateNode("AITemplateLoader", new JObject()
                {
                    ["model"] = g.LoadingModel,
                    ["keep_loaded"] = "disable"
                });
                g.LoadingModel = [aitLoad, 0];
            }
            if (g.UserInput.TryGet(T2IParamTypes.TorchCompile, out string compileMode) && compileMode != "Disabled")
            {
                string torchCompile = g.CreateNode("TorchCompileModel", new JObject()
                {
                    ["model"] = g.LoadingModel,
                    ["backend"] = compileMode
                });
                g.LoadingModel = [torchCompile, 0];
            }
        }, -3);
        #endregion
        #region Base Image
        AddStep(g =>
        {
            // TODO: if g.IsAudioModel(), InitAudio?
            if (!g.IsAudioModel() && g.UserInput.TryGet(T2IParamTypes.InitImage, out Image img))
            {
                string maskImageNode = null;
                if (g.UserInput.TryGet(T2IParamTypes.MaskImage, out Image mask))
                {
                    WGNodeData maskNode = g.LoadImage(mask, "${maskimage}", true);
                    maskImageNode = g.CreateNode("ImageToMask", new JObject()
                    {
                        ["image"] = maskNode.Path,
                        ["channel"] = "red"
                    });
                    g.EnableDifferential();
                    if (g.UserInput.TryGet(T2IParamTypes.MaskGrow, out int growAmount))
                    {
                        maskImageNode = g.CreateNode("SwarmMaskGrow", new JObject()
                        {
                            ["mask"] = NodePath(maskImageNode, 0),
                            ["grow"] = growAmount,
                        });
                    }
                    if (g.UserInput.TryGet(T2IParamTypes.MaskBlur, out int blurAmount))
                    {
                        maskImageNode = g.CreateNode("SwarmMaskBlur", new JObject()
                        {
                            ["mask"] = NodePath(maskImageNode, 0),
                            ["blur_radius"] = blurAmount,
                            ["sigma"] = 1.0
                        });
                    }
                    g.FinalMask = [maskImageNode, 0];
                }
                g.BasicInputImage = g.LoadImage(img, "${initimage}", true, "15");
                g.CurrentMedia = g.BasicInputImage;
                JArray currentMask = g.FinalMask;
                if (g.UserInput.TryGet(T2IParamTypes.InitImageNoise, out double initNoise))
                {
                    JObject noiseInput = new()
                    {
                        ["image"] = g.BasicInputImage.Path,
                        ["amount"] = initNoise,
                        ["seed"] = g.UserInput.Get(T2IParamTypes.Seed, 0) + 327
                    };
                    if (currentMask is not null)
                    {
                        // cut the edges of a blurred mask back to make recompositing cleaner
                        string thresholded = g.CreateNode("SwarmMaskThreshold", new JObject()
                        {
                            ["mask"] = currentMask,
                            ["min"] = 0.5,
                            ["max"] = 1.0
                        });
                        noiseInput["mask"] = NodePath(thresholded, 0);
                    }
                    string noised = g.CreateNode("SwarmImageNoise", noiseInput);
                    g.BasicInputImage = g.BasicInputImage.WithPath([noised, 0]);
                }
                if (currentMask is not null)
                {
                    if (g.UserInput.TryGet(T2IParamTypes.MaskShrinkGrow, out int shrinkGrow))
                    {
                        g.MaskShrunkInfo = g.CreateImageMaskCrop(g.FinalMask, g.BasicInputImage.Path, shrinkGrow, g.CurrentVae.Path, g.FinalLoadedModel);
                        currentMask = [g.MaskShrunkInfo.CroppedMask, 0];
                        // TODO: proper width/height wrangling
                        g.CurrentMedia = g.CurrentMedia.WithPath([g.MaskShrunkInfo.MaskedLatent, 0], WGNodeData.DT_LATENT_IMAGE);
                    }
                    else
                    {
                        JArray masked = g.DoMaskedVAEEncode(g.CurrentVae.Path, g.BasicInputImage.Path, currentMask, "5");
                        g.CurrentMedia = g.CurrentMedia.WithPath(masked, WGNodeData.DT_LATENT_IMAGE);
                    }
                }
                else
                {
                    g.CurrentMedia = g.BasicInputImage.EncodeToLatent(g.CurrentVae, "5");
                }
                if (g.UserInput.TryGet(T2IParamTypes.UnsamplerPrompt, out string unprompt))
                {
                    int steps = g.UserInput.Get(T2IParamTypes.Steps);
                    int startStep = 0;
                    if (g.UserInput.TryGet(T2IParamTypes.InitImageCreativity, out double creativity))
                    {
                        startStep = (int)Math.Round(steps * (1 - creativity));
                    }
                    JArray posCond = g.CreateConditioning(unprompt, g.CurrentTextEnc.Path, g.FinalLoadedModel, true);
                    JArray negCond = g.CreateConditioning(g.UserInput.Get(T2IParamTypes.NegativePrompt, ""), g.CurrentTextEnc.Path, g.FinalLoadedModel, false);
                    g.CurrentMedia = g.CurrentMedia.AsLatentImage(g.CurrentVae);
                    string unsampler = g.CreateNode("SwarmUnsampler", new JObject()
                    {
                        ["model"] = g.CurrentModel.Path,
                        ["steps"] = steps,
                        ["sampler_name"] = g.UserInput.Get(ComfyUIBackendExtension.SamplerParam, "euler"),
                        ["scheduler"] = g.UserInput.Get(ComfyUIBackendExtension.SchedulerParam, "normal"),
                        ["positive"] = posCond,
                        ["negative"] = negCond,
                        ["latent_image"] = g.CurrentMedia.Path,
                        ["start_at_step"] = startStep,
                        ["previews"] = g.UserInput.Get(T2IParamTypes.NoPreviews) ? "none" : "default"
                    });
                    g.CurrentMedia = g.CurrentMedia.WithPath([unsampler, 0]);
                    g.MainSamplerAddNoise = false;
                }
                if (g.UserInput.TryGet(T2IParamTypes.BatchSize, out int batchSize) && batchSize > 1)
                {
                    g.CurrentMedia = g.CurrentMedia.AsLatentImage(g.CurrentVae);
                    string batchNode = g.CreateNode("RepeatLatentBatch", new JObject()
                    {
                        ["samples"] = g.CurrentMedia.Path,
                        ["amount"] = batchSize
                    });
                    g.CurrentMedia = g.CurrentMedia.WithPath([batchNode, 0]);
                }
                else
                {
                    batchSize = 1;
                }
                if (g.UserInput.TryGet(T2IParamTypes.InitImageResetToNorm, out double resetFactor))
                {
                    g.CurrentMedia = g.CurrentMedia.AsLatentImage(g.CurrentVae);
                    string emptyImg = g.CreateEmptyImage(g.UserInput.GetImageWidth(), g.UserInput.GetImageHeight(), batchSize);
                    if (g.Features.Contains("comfy_latent_blend_masked") && currentMask is not null)
                    {
                        string blended = g.CreateNode("SwarmLatentBlendMasked", new JObject()
                        {
                            ["samples0"] = g.CurrentMedia.Path,
                            ["samples1"] = NodePath(emptyImg, 0),
                            ["mask"] = currentMask,
                            ["blend_factor"] = resetFactor
                        });
                        g.CurrentMedia = g.CurrentMedia.WithPath([blended, 0]);
                    }
                    else
                    {
                        string emptyMultiplied = g.CreateNode("LatentMultiply", new JObject()
                        {
                            ["samples"] = NodePath(emptyImg, 0),
                            ["multiplier"] = resetFactor
                        });
                        string originalMultiplied = g.CreateNode("LatentMultiply", new JObject()
                        {
                            ["samples"] = g.CurrentMedia.Path,
                            ["multiplier"] = 1 - resetFactor
                        });
                        string added = g.CreateNode("LatentAdd", new JObject()
                        {
                            ["samples1"] = NodePath(emptyMultiplied, 0),
                            ["samples2"] = NodePath(originalMultiplied, 0)
                        });
                        g.CurrentMedia = g.CurrentMedia.WithPath([added, 0]);
                    }
                }
                if (g.IsLTXV2())
                {
                    if (g.CurrentMedia.DataType == WGNodeData.DT_VIDEO)
                    {
                        string frameCountNode = g.CreateNode("SwarmCountFrames", new JObject()
                        {
                            ["image"] = g.CurrentMedia.AsRawImage(g.CurrentVae).Path
                        });
                        string emptyAudio = g.CreateNode("LTXVEmptyLatentAudio", new JObject()
                        {
                            ["batch_size"] = batchSize,
                            ["frames_number"] = NodePath(frameCountNode, 0),
                            ["frame_rate"] = g.CurrentMedia.FPS ?? g.UserInput.Get(T2IParamTypes.VideoFPS, 24),
                            ["audio_vae"] = g.CurrentAudioVae.Path
                        });
                        g.CurrentMedia = g.CurrentMedia.Duplicate();
                        g.CurrentMedia.Frames = null;
                        g.CurrentMedia.AttachedAudio = new([emptyAudio, 0], g, WGNodeData.DT_LATENT_AUDIO, g.CurrentCompat());
                    }
                }
            }
            else
            {
                g.CurrentMedia = g.EmptyImage(g.UserInput.GetImageWidth(), g.UserInput.GetImageHeight(), g.UserInput.Get(T2IParamTypes.BatchSize, 1), "5");
            }
            if (g.UserInput.TryGet(T2IParamTypes.VideoAudioInput, out AudioFile audioData))
            {
                string audioNode = g.CreateAudioLoadNode(audioData, "${videoaudioinput}");
                g.CurrentMedia.AttachedAudio = new WGNodeData([audioNode, 0], g, WGNodeData.DT_AUDIO, g.CurrentCompat());
            }
        }, -9);
        #endregion
        #region Positive Prompt
        AddStep(g =>
        {
            g.FinalPrompt = g.CreateConditioning(g.UserInput.Get(T2IParamTypes.Prompt), g.CurrentTextEnc.Path, g.UserInput.Get(T2IParamTypes.Model), true, "6");
        }, -8);
        #endregion
        #region ReVision/UnCLIP/IPAdapter
        AddStep(g =>
        {
            if (g.UserInput.TryGet(T2IParamTypes.PromptImages, out List<Image> images) && images.Any())
            {
                if (g.UserInput.TryGet(ComfyUIBackendExtension.UseStyleModel, out string styleModelName))
                {
                    string clipVis = g.RequireVisionModel("sigclip_vision_patch14_384.safetensors", "https://huggingface.co/Comfy-Org/sigclip_vision_384/resolve/main/sigclip_vision_patch14_384.safetensors", "1fee501deabac72f0ed17610307d7131e3e9d1e838d0363aa3c2b97a6e03fb33", T2IParamTypes.ClipVisionModel);
                    string styleModelClipLoader = g.CreateNode("CLIPVisionLoader", new JObject()
                    {
                        ["clip_name"] = clipVis
                    });
                    string styleModelLoader = g.CreateNode("StyleModelLoader", new JObject()
                    {
                        ["style_model_name"] = styleModelName
                    });
                    for (int i = 0; i < images.Count; i++)
                    {
                        WGNodeData imageLoader = g.LoadImage(images[i], "${promptimages." + i + "}", false);
                        string encoded = g.CreateNode("CLIPVisionEncode", new JObject()
                        {
                            ["clip_vision"] = NodePath(styleModelClipLoader, 0),
                            ["image"] = imageLoader.Path,
                            ["crop"] = "none"
                        });
                        string styled = g.CreateNode("StyleModelApply", new JObject()
                        {
                            ["conditioning"] = g.FinalPrompt,
                            ["clip_vision_output"] = NodePath(encoded, 0),
                            ["style_model"] = NodePath(styleModelLoader, 0),
                            ["strength_type"] = "multiply",
                            ["strength"] = g.UserInput.Get(ComfyUIBackendExtension.StyleModelMultiplyStrength, 1)
                        });
                        if (g.UserInput.TryGet(ComfyUIBackendExtension.StyleModelMergeStrength, out double mergeStrength) && mergeStrength < 1)
                        {
                            styled = g.CreateNode("ConditioningAverage", new JObject()
                            {
                                ["conditioning_to"] = NodePath(styled, 0),
                                ["conditioning_from"] = g.FinalPrompt,
                                ["conditioning_to_strength"] = mergeStrength
                            });
                        }
                        if (g.UserInput.TryGet(ComfyUIBackendExtension.StyleModelApplyStart, out double applyAt) && applyAt > 0)
                        {
                            string cond1 = g.CreateNode("ConditioningSetTimestepRange", new JObject()
                            {
                                ["conditioning"] = g.FinalPrompt,
                                ["start"] = 0,
                                ["end"] = applyAt
                            });
                            string cond2 = g.CreateNode("ConditioningSetTimestepRange", new JObject()
                            {
                                ["conditioning"] = NodePath(styled, 0),
                                ["start"] = applyAt,
                                ["end"] = 1
                            });
                            string combined = g.CreateNode("ConditioningCombine", new JObject()
                            {
                                ["conditioning_1"] = NodePath(cond1, 0),
                                ["conditioning_2"] = NodePath(cond2, 0),
                            });
                            g.FinalPrompt = [combined, 0];
                        }
                        else
                        {
                            g.FinalPrompt = [styled, 0];
                        }
                    }
                }
                string visionLoaderId = null;
                string getVisionLoader()
                {
                    if (visionLoaderId is not null)
                    {
                        return visionLoaderId;
                    }
                    string visModelName = "clip_vision_g.safetensors";
                    visModelName = g.RequireVisionModel(visModelName, "https://huggingface.co/stabilityai/control-lora/resolve/main/revision/clip_vision_g.safetensors", "9908329b3ead722a693ea400fab1d7c9ec91d6736fd194a94d20d793457f9c2e", T2IParamTypes.ClipVisionModel);
                    visionLoaderId = g.CreateNode("CLIPVisionLoader", new JObject()
                    {
                        ["clip_name"] = visModelName
                    });
                    return visionLoaderId;
                }
                double revisionStrength = g.UserInput.Get(T2IParamTypes.ReVisionStrength, 0);
                if (revisionStrength > 0)
                {
                    bool autoZero = g.UserInput.Get(T2IParamTypes.RevisionZeroPrompt, false);
                    if ((g.UserInput.TryGet(T2IParamTypes.Prompt, out string promptText) && string.IsNullOrWhiteSpace(promptText)) || autoZero)
                    {
                        string zeroed = g.CreateNode("ConditioningZeroOut", new JObject()
                        {
                            ["conditioning"] = g.FinalPrompt
                        });
                        g.FinalPrompt = [zeroed, 0];
                    }
                    if ((g.UserInput.TryGet(T2IParamTypes.NegativePrompt, out string negPromptText) && string.IsNullOrWhiteSpace(negPromptText)) || autoZero)
                    {
                        string zeroed = g.CreateNode("ConditioningZeroOut", new JObject()
                        {
                            ["conditioning"] = g.FinalNegativePrompt
                        });
                        g.FinalNegativePrompt = [zeroed, 0];
                    }
                    if (!g.UserInput.TryGet(T2IParamTypes.Model, out T2IModel model) || model.ModelClass is null ||
                        (model.ModelClass.CompatClass?.ID != "stable-diffusion-xl-v1"/* && model.ModelClass.CompatClass?.ID != "stable-diffusion-v3-medium"*/))
                    {
                        throw new SwarmUserErrorException($"Model type must be SDXL for ReVision (currently is {model?.ModelClass?.Name ?? "Unknown"}). Set ReVision Strength to 0 if you just want IP-Adapter.");
                    }
                    for (int i = 0; i < images.Count; i++)
                    {
                        WGNodeData imageLoader = g.LoadImage(images[i], "${promptimages." + i + "}", false);
                        string encoded = g.CreateNode("CLIPVisionEncode", new JObject()
                        {
                            ["clip_vision"] = NodePath($"{getVisionLoader()}", 0),
                            ["image"] = imageLoader.Path,
                            ["crop"] = "none"
                        });
                        string unclipped = g.CreateNode("unCLIPConditioning", new JObject()
                        {
                            ["conditioning"] = g.FinalPrompt,
                            ["clip_vision_output"] = NodePath(encoded, 0),
                            ["strength"] = revisionStrength,
                            ["noise_augmentation"] = 0
                        });
                        g.FinalPrompt = [unclipped, 0];
                    }
                }
                if (g.UserInput.Get(T2IParamTypes.UseReferenceOnly, false))
                {
                    WGNodeData firstImgNode = g.LoadImage(images[0], "${promptimages.0}", true);
                    WGNodeData lastEncoded = firstImgNode.AsLatentImage(g.CurrentVae);
                    for (int i = 1; i < images.Count; i++)
                    {
                        WGNodeData newImgNode = g.LoadImage(images[i], "${promptimages." + i + "}", true);
                        WGNodeData newEncoded = newImgNode.AsLatentImage(g.CurrentVae);
                        string newBatched = g.CreateNode("LatentBatch", new JObject()
                        {
                            ["samples1"] = lastEncoded.Path,
                            ["samples2"] = newEncoded.Path
                        });
                        lastEncoded = lastEncoded.WithPath([newBatched, 0]);
                    }
                    g.CurrentMedia = g.CurrentMedia.AsLatentImage(g.CurrentVae);
                    string referencedModel = g.CreateNode("SwarmReferenceOnly", new JObject()
                    {
                        ["model"] = g.CurrentModel.Path,
                        ["reference"] = lastEncoded.Path,
                        ["latent"] = g.CurrentMedia.Path
                    });
                    g.CurrentModel = g.CurrentModel.WithPath([referencedModel, 0]);
                    g.CurrentMedia = g.CurrentMedia.WithPath([referencedModel, 1]);
                    g.DefaultPreviews = "second";
                }
                if (g.UserInput.TryGet(ComfyUIBackendExtension.UseIPAdapterForRevision, out string ipAdapter) && ipAdapter != "None")
                {
                    string getIPAvisionLoader()
                    {
                        string ipAdapterVisionLoader = getVisionLoader();
                        if (!g.Features.Contains("cubiqipadapterunified"))
                        {
                            if ((ipAdapter.Contains("sd15") && !ipAdapter.Contains("vit-G")) || ipAdapter.Contains("vit-h"))
                            {
                                string targetName = "clip_vision_h.safetensors";
                                targetName = g.RequireVisionModel(targetName, "https://huggingface.co/h94/IP-Adapter/resolve/main/models/image_encoder/model.safetensors", "6ca9667da1ca9e0b0f75e46bb030f7e011f44f86cbfb8d5a36590fcd7507b030", T2IParamTypes.ClipVisionModel);
                                ipAdapterVisionLoader = g.CreateNode("CLIPVisionLoader", new JObject()
                                {
                                    ["clip_name"] = targetName
                                });
                            }
                        }
                        return ipAdapterVisionLoader;
                    }
                    if (g.Features.Contains("cubiqipadapterunified"))
                    {
                        g.RequireVisionModel("CLIP-ViT-H-14-laion2B-s32B-b79K.safetensors", "https://huggingface.co/h94/IP-Adapter/resolve/main/models/image_encoder/model.safetensors", "6ca9667da1ca9e0b0f75e46bb030f7e011f44f86cbfb8d5a36590fcd7507b030");
                        g.RequireVisionModel("CLIP-ViT-bigG-14-laion2B-39B-b160k.safetensors", "https://huggingface.co/h94/IP-Adapter/resolve/main/sdxl_models/image_encoder/model.safetensors", "657723e09f46a7c3957df651601029f66b1748afb12b419816330f16ed45d64d");
                    }
                    WGNodeData lastImage = g.LoadImage(images[0], "${promptimages.0}", false);
                    for (int i = 1; i < images.Count; i++)
                    {
                        WGNodeData newImg = g.LoadImage(images[i], "${promptimages." + i + "}", false);
                        string batched = g.CreateNode("ImageBatch", new JObject()
                        {
                            ["image1"] = lastImage.Path,
                            ["image2"] = newImg.Path
                        });
                        lastImage = lastImage.WithPath([batched, 0]);
                    }
                    if (g.Features.Contains("cubiqipadapterunified"))
                    {
                        string presetLow = ipAdapter.ToLowerFast();
                        bool isXl = g.CurrentCompatClass() == "stable-diffusion-xl-v1";
                        void requireIPAdapterModel(string name, string url, string hash)
                        {
                            if (IPAdapterModelsValid.ContainsKey(name))
                            {
                                return;
                            }
                            string filePath = Utilities.CombinePathWithAbsolute(Program.ServerSettings.Paths.ActualModelRoot, $"ipadapter/{name}");
                            g.DownloadModel(name, filePath, url, hash);
                            IPAdapterModelsValid.TryAdd(name, name);
                        }
                        void requireLora(string name, string url, string hash)
                        {
                            if (IPAdapterModelsValid.ContainsKey($"LORA-{name}"))
                            {
                                return;
                            }
                            string filePath = Utilities.CombinePathWithAbsolute(Program.ServerSettings.Paths.ActualModelRoot, Program.ServerSettings.Paths.SDLoraFolder.Split(';')[0], $"ipadapter/{name}");
                            g.DownloadModel(name, filePath, url, hash);
                            IPAdapterModelsValid.TryAdd($"LORA-{name}", name);
                        }
                        // IPAdapter model links @ https://github.com/cubiq/ComfyUI_IPAdapter_plus?tab=readme-ov-file#installation
                        // required model for any given type @ https://github.com/cubiq/ComfyUI_IPAdapter_plus/blob/main/utils.py#L29
                        if (presetLow.StartsWith("file:"))
                        {
                            // no autodownload
                        }
                        else if (presetLow.StartsWith("light"))
                        {
                            if (isXl) { throw new SwarmUserErrorException("IP-Adapter light model is not supported for SDXL"); }
                            else { requireIPAdapterModel("sd15_light_v11.bin", "https://huggingface.co/h94/IP-Adapter/resolve/main/models/ip-adapter_sd15_light_v11.bin", "350b63a57847c163e2e984b01090f85ffe60eaae20f32b2b2c9e1ccc7ddd972b"); }
                        }
                        else if (presetLow.StartsWith("standard"))
                        {
                            if (isXl) { requireIPAdapterModel("ip-adapter_sdxl_vit-h.safetensors", "https://huggingface.co/h94/IP-Adapter/resolve/main/sdxl_models/ip-adapter_sdxl_vit-h.safetensors", "ebf05d918348aec7abb02a5e9ecef77e0aaea6914a5c4ea13f50d45eb1681831"); }
                            else { requireIPAdapterModel("ip-adapter_sd15.safetensors", "https://huggingface.co/h94/IP-Adapter/resolve/main/models/ip-adapter_sd15.safetensors", "289b45f16d043d0bf542e45831f971dcdaabe18b656f11e86d9dfba7e9ee3369"); }
                        }
                        else if (presetLow.StartsWith("vit-g"))
                        {
                            if (isXl) { requireIPAdapterModel("ip-adapter_sdxl.safetensors", "https://huggingface.co/h94/IP-Adapter/resolve/main/sdxl_models/ip-adapter_sdxl.safetensors", "ba1002529e783604c5f326d49f0122025392d1d20ac8d573b3eeb3e6dea4ebb6"); }
                            else { requireIPAdapterModel("ip-adapter_sd15_vit-G.safetensors", "https://huggingface.co/h94/IP-Adapter/resolve/main/models/ip-adapter_sd15_vit-G.safetensors", "a26f736af07bb341a83dfea23713531d0575760e8ed947c68cb31a4c62d9c90b"); }
                        }
                        else if (presetLow.StartsWith("plus ("))
                        {
                            if (isXl) { requireIPAdapterModel("ip-adapter-plus_sdxl_vit-h.safetensors", "https://huggingface.co/h94/IP-Adapter/resolve/main/sdxl_models/ip-adapter-plus_sdxl_vit-h.safetensors", "3f5062b8400c94b7159665b21ba5c62acdcd7682262743d7f2aefedef00e6581"); }
                            else { requireIPAdapterModel("ip-adapter-plus_sd15.safetensors", "https://huggingface.co/h94/IP-Adapter/resolve/main/models/ip-adapter-plus_sd15.safetensors", "a1c250be40455cc61a43da1201ec3f1edaea71214865fb47f57927e06cbe4996"); }
                        }
                        else if (presetLow.StartsWith("plus face"))
                        {
                            if (isXl) { requireIPAdapterModel("ip-adapter-plus-face_sdxl_vit-h.safetensors", "https://huggingface.co/h94/IP-Adapter/resolve/main/sdxl_models/ip-adapter-plus-face_sdxl_vit-h.safetensors", "677ad8860204f7d0bfba12d29e6c31ded9beefdf3e4bbd102518357d31a292c1"); }
                            else { requireIPAdapterModel("ip-adapter-plus-face_sd15.safetensors", "https://huggingface.co/h94/IP-Adapter/resolve/main/models/ip-adapter-plus-face_sd15.safetensors", "1c9edc21af6f737dc1d6e0e734190e976cfacf802d6b024b77aa3be922f7569b"); }
                        }
                        else if (presetLow.StartsWith("full"))
                        {
                            if (isXl) { throw new SwarmUserErrorException("IP-Adapter full face model is not supported for SDXL"); }
                            else { requireIPAdapterModel("full_face_sd15.safetensors", "https://huggingface.co/h94/IP-Adapter/resolve/main/models/ip-adapter-full-face_sd15.safetensors", "f4a17fb643bf876235a45a0e87a49da2855be6584b28ca04c62a97ab5ff1c6f3"); }
                        }
                        else if (presetLow == "faceid")
                        {
                            if (isXl)
                            {
                                requireIPAdapterModel("ip-adapter-faceid_sdxl.bin", "https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid_sdxl.bin", "f455fed24e207c878ec1e0466b34a969d37bab857c5faa4e8d259a0b4ff63d7e");
                                requireLora("ip-adapter-faceid_sdxl_lora.safetensors", "https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid_sdxl_lora.safetensors", "4fcf93d6e8dc8dd18f5f9e51c8306f369486ed0aa0780ade9961308aff7f0d64");
                            }
                            else
                            {
                                requireIPAdapterModel("ip-adapter-faceid_sd15.bin", "https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid_sd15.bin", "201344e22e6f55849cf07ca7a6e53d8c3b001327c66cb9710d69fd5da48a8da7");
                                requireLora("ip-adapter-faceid_sd15_lora.safetensors", "https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid_sd15_lora.safetensors", "70699f0dbfadd47de1f81d263cf4c86bd4b7271d841304af9b340b3a7f38e86a");
                            }
                        }
                        else if (presetLow.StartsWith("faceid plus -"))
                        {
                            if (isXl) { throw new SwarmUserErrorException("IP-Adapter FaceID plus model is not supported for SDXL"); }
                            else
                            {
                                requireIPAdapterModel("ip-adapter-faceid-plus_sd15.bin", "https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid-plus_sd15.bin", "252fb53e0d018489d9e7f9b9e2001a52ff700e491894011ada7cfb471e0fadf2");
                                requireLora("ip-adapter-faceid-plus_sd15_lora.safetensors", "https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid-plus_sd15_lora.safetensors", "3f00341d11e5e7b5aadf63cbdead09ef82eb28669156161cf1bfc2105d4ff1cd");
                            }
                        }
                        else if (presetLow.StartsWith("faceid plus v2"))
                        {
                            if (isXl)
                            {
                                requireIPAdapterModel("ip-adapter-faceid-plusv2_sdxl.bin", "https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid-plusv2_sdxl.bin", "c6945d82b543700cc3ccbb98d363b837e9c596281607857c74b713a876daf5fb");
                                requireLora("ip-adapter-faceid-plusv2_sdxl_lora.safetensors", "https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid-plusv2_sdxl_lora.safetensors", "f24b4bb2dad6638a09c00f151cde84991baf374409385bcbab53c1871a30cb7b");
                            }
                            else
                            {
                                requireIPAdapterModel("ip-adapter-faceid-plusv2_sd15.bin", "https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid-plusv2_sd15.bin", "26d0d86a1d60d6cc811d3b8862178b461e1eeb651e6fe2b72ba17aa95411e313");
                                requireLora("ip-adapter-faceid-plusv2_sd15_lora.safetensors", "https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid-plusv2_sd15_lora.safetensors", "8abff87a15a049f3e0186c2e82c1c8e77783baf2cfb63f34c412656052eb57b0");
                            }
                        }
                        else if (presetLow.StartsWith("faceid portrait unnorm"))
                        {
                            if (isXl) { requireIPAdapterModel("ip-adapter-faceid-portrait_sdxl_unnorm.bin", "https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid-portrait_sdxl_unnorm.bin", "220bb86e205393a3d0411631cb473caddbf35fd371be2905ca9008818170db55"); }
                            else { throw new SwarmUserErrorException("IP-Adapter FaceID Portrait UnNorm model is only supported for SDXL"); }
                        }
                        else if (presetLow.StartsWith("faceid portrait"))
                        {
                            if (isXl) { requireIPAdapterModel("ip-adapter-faceid-portrait_sdxl.bin", "https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid-portrait_sdxl.bin", "5631ce7824cdafd2db37c5e85b985730a95ff59c5b4fc80c2b79b0bee5711512"); }
                            else { requireIPAdapterModel("ip-adapter-faceid-portrait-v11_sd15.bin", "https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid-portrait-v11_sd15.bin", "a48cb4f89ed18e02c6000f65aa9efec452e87eaed4a1bc9fcf4a460c8d0e3bc6"); }
                        }
                        string ipAdapterLoader;
                        if (presetLow.StartsWith("file:"))
                        {
                            ipAdapterLoader = g.CreateNode("IPAdapterModelLoader", new JObject()
                            {
                                ["ipadapter_file"] = ipAdapter["file:".Length..]
                            });
                        }
                        else if (presetLow.StartsWith("faceid"))
                        {
                            ipAdapterLoader = g.CreateNode("IPAdapterUnifiedLoaderFaceID", new JObject()
                            {
                                ["model"] = g.CurrentModel.Path,
                                ["preset"] = ipAdapter,
                                ["lora_strength"] = 0.6,
                                ["provider"] = "CPU"
                            });
                        }
                        else
                        {
                            ipAdapterLoader = g.CreateNode("IPAdapterUnifiedLoader", new JObject()
                            {
                                ["model"] = g.CurrentModel.Path,
                                ["preset"] = ipAdapter
                            });
                        }
                        double ipAdapterStart = g.UserInput.Get(ComfyUIBackendExtension.IPAdapterStart, 0.0);
                        double ipAdapterEnd = g.UserInput.Get(ComfyUIBackendExtension.IPAdapterEnd, 1.0);
                        if (ipAdapterStart >= ipAdapterEnd)
                        {
                            throw new SwarmUserErrorException($"IP-Adapter Start must be less than IP-Adapter End.");
                        }
                        if (presetLow.StartsWith("file:"))
                        {
                            string weightType = g.UserInput.Get(ComfyUIBackendExtension.IPAdapterWeightType, "linear");
                            if (weightType == "standard") { weightType = "linear"; }
                            else if (weightType == "prompt is more important") { weightType = "ease out"; }
                            string ipAdapterNode = g.CreateNode("IPAdapterAdvanced", new JObject()
                            {
                                ["model"] = g.CurrentModel.Path,
                                ["ipadapter"] = NodePath(ipAdapterLoader, 0),
                                ["image"] = lastImage.Path,
                                ["weight"] = g.UserInput.Get(ComfyUIBackendExtension.IPAdapterWeight, 1),
                                ["start_at"] = ipAdapterStart,
                                ["end_at"] = ipAdapterEnd,
                                ["weight_type"] = weightType,
                                ["combine_embeds"] = "concat",
                                ["embeds_scaling"] = "V only",
                                ["clip_vision"] = NodePath(getVisionLoader(), 0)
                            });
                            g.CurrentModel = g.CurrentModel.WithPath([ipAdapterNode, 0]);
                        }
                        else
                        {
                            string ipAdapterNode = g.CreateNode("IPAdapter", new JObject()
                            {
                                ["model"] = NodePath(ipAdapterLoader, 0),
                                ["ipadapter"] = NodePath(ipAdapterLoader, 1),
                                ["image"] = lastImage.Path,
                                ["weight"] = g.UserInput.Get(ComfyUIBackendExtension.IPAdapterWeight, 1),
                                ["start_at"] = ipAdapterStart,
                                ["end_at"] = ipAdapterEnd,
                                ["weight_type"] = g.UserInput.Get(ComfyUIBackendExtension.IPAdapterWeightType, "standard")
                            });
                            g.CurrentModel = g.CurrentModel.WithPath([ipAdapterNode, 0]);
                        }
                    }
                    else if (g.Features.Contains("cubiqipadapter"))
                    {
                        string ipAdapterLoader = g.CreateNode("IPAdapterModelLoader", new JObject()
                        {
                            ["ipadapter_file"] = ipAdapter
                        });
                        string ipAdapterNode = g.CreateNode("IPAdapterApply", new JObject()
                        {
                            ["ipadapter"] = NodePath(ipAdapterLoader, 0),
                            ["model"] = g.CurrentModel.Path,
                            ["image"] = lastImage.Path,
                            ["clip_vision"] = NodePath(getIPAvisionLoader(), 0),
                            ["weight"] = g.UserInput.Get(ComfyUIBackendExtension.IPAdapterWeight, 1),
                            ["noise"] = 0,
                            ["weight_type"] = "original"
                        });
                        g.CurrentModel = g.CurrentModel.WithPath([ipAdapterNode, 0]);
                    }
                    else
                    {
                        string ipAdapterNode = g.CreateNode("IPAdapter", new JObject()
                        {
                            ["model"] = g.CurrentModel.Path,
                            ["image"] = lastImage.Path,
                            ["clip_vision"] = NodePath(getIPAvisionLoader(), 0),
                            ["weight"] = g.UserInput.Get(ComfyUIBackendExtension.IPAdapterWeight, 1),
                            ["model_name"] = ipAdapter,
                            ["dtype"] = "fp16" // TODO: ...???
                        });
                        g.CurrentModel = g.CurrentModel.WithPath([ipAdapterNode, 0]);
                    }
                }
            }
        }, -7);
        #endregion
        #region Negative Prompt
        AddStep(g =>
        {
            g.FinalNegativePrompt = g.CreateConditioning(g.UserInput.Get(T2IParamTypes.NegativePrompt, ""), g.CurrentTextEnc.Path, g.UserInput.Get(T2IParamTypes.Model), false, "7");
        }, -7);
        #endregion
        #region ControlNet
        AddStep(g =>
        {
            WGNodeData firstImageNode = null;
            for (int i = 0; i < 3; i++)
            {
                T2IParamTypes.ControlNetParamHolder controlnetParams = T2IParamTypes.Controlnets[i];
                if (g.UserInput.TryGet(controlnetParams.Strength, out double controlStrength))
                {
                    string imageInput = "${" + controlnetParams.Image.Type.ID + "}";
                    WGNodeData imageNodeActual = null;
                    if (g.UserInput.TryGet(controlnetParams.Image, out Image img))
                    {
                        imageNodeActual = g.LoadImage(img, imageInput, true);
                    }
                    else
                    {
                        if (i == 0 ? g.BasicInputImage is null : firstImageNode is null)
                        {
                            Logs.Verbose($"Following error relates to parameters: {g.UserInput.ToJSON().ToDenseDebugString()}");
                            throw new SwarmUserErrorException("Must specify either a ControlNet Image, or Init image. Or turn off ControlNet if not wanted.");
                        }
                        if (i == 0)
                        {
                            firstImageNode = g.BasicInputImage;
                            if (g.MaskShrunkInfo.ScaledImage is not null)
                            {
                                firstImageNode = new([g.MaskShrunkInfo.ScaledImage, 0], g, WGNodeData.DT_IMAGE, g.CurrentCompat());
                            }
                        }
                        imageNodeActual = firstImageNode;
                    }
                    if (i == 0)
                    {
                        firstImageNode = imageNodeActual;
                    }
                    T2IModel controlModel = g.UserInput.Get(controlnetParams.Model, null);
                    if (!g.UserInput.TryGet(ComfyUIBackendExtension.ControlNetPreprocessorParams[i], out string preprocessor))
                    {
                        preprocessor = "none";
                        string wantedPreproc = controlModel?.Metadata?.Preprocessor;
                        string cnName = $"{controlModel?.Name}{controlModel?.RawFilePath.Replace('\\', '/').AfterLast('/')}".ToLowerFast();
                        if (string.IsNullOrWhiteSpace(wantedPreproc))
                        {
                            if (cnName.Contains("canny")) { wantedPreproc = "canny"; }
                            else if (cnName.Contains("depth") || cnName.Contains("midas")) { wantedPreproc = "depth"; }
                            else if (cnName.Contains("sketch")) { wantedPreproc = "sketch"; }
                            else if (cnName.Contains("scribble")) { wantedPreproc = "scribble"; }
                            else if (cnName.Contains("pose")) { wantedPreproc = "pose"; }
                        }
                        if (string.IsNullOrWhiteSpace(wantedPreproc))
                        {
                            Logs.Verbose($"No wanted preprocessor, and '{cnName}' doesn't imply any other option, skipping...");
                        }
                        else
                        {
                            string[] procs = [.. ComfyUIBackendExtension.ControlNetPreprocessors.Keys];
                            bool getBestFor(string phrase)
                            {
                                string result = procs.FirstOrDefault(m => m.ToLowerFast().Contains(phrase.ToLowerFast()));
                                if (result is not null)
                                {
                                    preprocessor = result;
                                    return true;
                                }
                                return false;
                            }
                            if (wantedPreproc == "depth")
                            {
                                if (!getBestFor("midas-depthmap") && !getBestFor("depthmap") && !getBestFor("depth") && !getBestFor("midas") && !getBestFor("zoe") && !getBestFor("leres"))
                                {
                                    throw new SwarmUserErrorException("No preprocessor found for depth - please install a Comfy extension that adds eg MiDaS depthmap preprocessors, or select 'none' if using a manual depthmap");
                                }
                            }
                            else if (wantedPreproc == "canny")
                            {
                                if (!getBestFor("cannyedge") && !getBestFor("canny"))
                                {
                                    preprocessor = "none";
                                }
                            }
                            else if (wantedPreproc == "sketch")
                            {
                                if (!getBestFor("sketch") && !getBestFor("lineart") && !getBestFor("scribble"))
                                {
                                    preprocessor = "none";
                                }
                            }
                            else if (wantedPreproc == "pose")
                            {
                                if (!getBestFor("openpose") && !getBestFor("pose"))
                                {
                                    preprocessor = "none";
                                }
                            }
                            else
                            {
                                Logs.Verbose($"Wanted preprocessor {wantedPreproc} unrecognized, skipping...");
                            }
                        }
                    }
                    if (preprocessor.ToLowerFast() != "none")
                    {
                        if (imageNodeActual.DataType == WGNodeData.DT_VIDEO && imageNodeActual.FPS is not null)
                        {
                            int fps = g.Text2VideoFPS();
                            string resampleNode = g.CreateNode("SwarmVideoResampleFPS", new JObject()
                            {
                                ["images"] = imageNodeActual.Path,
                                ["fps_in"] = imageNodeActual.FPS,
                                ["fps_out"] = fps,
                                ["method"] = "linear"
                            });
                            imageNodeActual = imageNodeActual.WithPath([resampleNode, 0]);
                            imageNodeActual.FPS = fps;
                        }
                        JArray preprocActual = g.CreatePreprocessor(preprocessor, imageNodeActual);
                        g.NodeHelpers["controlnet_preprocessor"] = $"{preprocActual[0]}";
                        imageNodeActual = imageNodeActual.WithPath(preprocActual);
                        string multipleOf8 = g.CreateNode("ResizeImageMaskNode", new JObject()
                        {
                            ["input"] = imageNodeActual.Path,
                            ["resize_type"] = "scale to multiple",
                            ["resize_type.multiple"] = 8,
                            ["scale_method"] = "lanczos"
                        });
                        imageNodeActual = imageNodeActual.WithPath([multipleOf8, 0]);
                        if (g.UserInput.Get(T2IParamTypes.ControlNetPreviewOnly))
                        {
                            g.CurrentMedia = imageNodeActual;
                            g.CurrentMedia.SaveOutput(g.CurrentVae, g.CurrentAudioVae, id: "9");
                            g.SkipFurtherSteps = true;
                            return;
                        }
                    }
                    else if (g.UserInput.Get(T2IParamTypes.ControlNetPreviewOnly))
                    {
                        throw new SwarmUserErrorException("Cannot preview a ControlNet preprocessor without any preprocessor enabled.");
                    }
                    if (controlModel is null)
                    {
                        throw new SwarmUserErrorException("Cannot use ControlNet without a model selected.");
                    }
                    if (imageNodeActual.DataType == WGNodeData.DT_VIDEO && !g.IsVideoModel())
                    {
                        string singleFrame = g.CreateNode("ImageFromBatch", new JObject()
                        {
                            ["image"] = imageNodeActual.Path,
                            ["batch_index"] = 0,
                            ["length"] = 1
                        });
                        imageNodeActual = imageNodeActual.WithPath([singleFrame, 0], WGNodeData.DT_IMAGE);
                    }
                    if (controlModel.ModelClass?.ID?.EndsWith("/control-diffpatch") ?? false)
                    {
                        string modelPatchLoader = g.CreateNode("ModelPatchLoader", new JObject()
                        {
                            ["name"] = controlModel.ToString(g.ModelFolderFormat)
                        });
                        string diffsynthNode = g.CreateNode("QwenImageDiffsynthControlnet", new JObject()
                        {
                            ["model"] = g.CurrentModel.Path,
                            ["model_patch"] = NodePath(modelPatchLoader, 0),
                            ["vae"] = g.CurrentVae.Path,
                            ["image"] = imageNodeActual.Path,
                            ["strength"] = controlStrength
                        });
                        g.CurrentModel = g.CurrentModel.WithPath([diffsynthNode, 0]);
                        continue;
                    }
                    string controlModelNode = g.CreateNode("ControlNetLoader", new JObject()
                    {
                        ["control_net_name"] = controlModel.ToString(g.ModelFolderFormat)
                    });
                    if (g.UserInput.TryGet(ComfyUIBackendExtension.ControlNetUnionTypeParams[i], out string unionType))
                    {
                        controlModelNode = g.CreateNode("SetUnionControlNetType", new JObject()
                        {
                            ["control_net"] = NodePath(controlModelNode, 0),
                            ["type"] = unionType
                        });
                    }
                    string applyNode;
                    if (controlModel.Metadata?.ModelClassType == "flux.1-dev/controlnet-alimamainpaint")
                    {
                        if (g.FinalMask is null)
                        {
                            throw new SwarmUserErrorException("Alimama Inpainting ControlNet requires a mask.");
                        }
                        applyNode = g.CreateNode("ControlNetInpaintingAliMamaApply", new JObject()
                        {
                            ["positive"] = g.FinalPrompt,
                            ["negative"] = g.FinalNegativePrompt,
                            ["control_net"] = NodePath(controlModelNode, 0),
                            ["vae"] = g.CurrentVae.Path,
                            ["image"] = imageNodeActual.Path,
                            ["mask"] = g.FinalMask,
                            ["strength"] = controlStrength,
                            ["start_percent"] = g.UserInput.Get(controlnetParams.Start, 0),
                            ["end_percent"] = g.UserInput.Get(controlnetParams.End, 1)
                        });
                    }
                    else if (g.IsSD3() || g.IsFlux() || g.IsAnyFlux2() || g.IsChroma() || g.IsQwenImage())
                    {
                        applyNode = g.CreateNode("ControlNetApplyAdvanced", new JObject()
                        {
                            ["positive"] = g.FinalPrompt,
                            ["negative"] = g.FinalNegativePrompt,
                            ["control_net"] = NodePath(controlModelNode, 0),
                            ["vae"] = g.CurrentVae.Path,
                            ["image"] = imageNodeActual.Path,
                            ["strength"] = controlStrength,
                            ["start_percent"] = g.UserInput.Get(controlnetParams.Start, 0),
                            ["end_percent"] = g.UserInput.Get(controlnetParams.End, 1)
                        });
                    }
                    else
                    {
                        applyNode = g.CreateNode("ControlNetApplyAdvanced", new JObject()
                        {
                            ["positive"] = g.FinalPrompt,
                            ["negative"] = g.FinalNegativePrompt,
                            ["control_net"] = NodePath(controlModelNode, 0),
                            ["image"] = imageNodeActual.Path,
                            ["strength"] = controlStrength,
                            ["start_percent"] = g.UserInput.Get(controlnetParams.Start, 0),
                            ["end_percent"] = g.UserInput.Get(controlnetParams.End, 1)
                        });
                    }
                    g.FinalPrompt = [applyNode, 0];
                    g.FinalNegativePrompt = [applyNode, 1];
                }
            }
            if (g.IsWanVace() && g.BasicInputImage is not null)
            {
                int width = g.UserInput.GetImageWidth();
                int height = g.UserInput.GetImageHeight();
                int frames = g.UserInput.Get(T2IParamTypes.Text2VideoFrames, 81);
                string vaceNode = g.CreateNode("WanVaceToVideo", new JObject()
                {
                    ["positive"] = g.FinalPrompt,
                    ["negative"] = g.FinalNegativePrompt,
                    ["vae"] = g.CurrentVae.Path,
                    ["reference_image"] = g.BasicInputImage.Path,
                    ["width"] = width,
                    ["height"] = height,
                    ["length"] = frames,
                    ["batch_size"] = g.UserInput.Get(T2IParamTypes.BatchSize, 1),
                    ["strength"] = 1 // TODO: ? Maybe hijack and redirect the creativity param?
                });
                g.FinalPrompt = [vaceNode, 0];
                g.FinalNegativePrompt = [vaceNode, 1];
                g.CurrentMedia = new WGNodeData([vaceNode, 2], g, WGNodeData.DT_LATENT_VIDEO, g.CurrentCompat()) { Width = width, Height = height, Frames = frames };
                g.FinalTrimLatent = [vaceNode, 3];
            }
            if (g.IsLTXV2() && g.UserInput.TryGet(T2IParamTypes.VideoAudioReference, out AudioFile audio))
            {
                string audioNode = g.CreateAudioLoadNode(audio, "${videoaudioinput}");
                string refNode = g.CreateNode("LTXVReferenceAudio", new JObject()
                {
                    ["model"] = g.CurrentModel.Path,
                    ["positive"] = g.FinalPrompt,
                    ["negative"] = g.FinalNegativePrompt,
                    ["reference_audio"] = NodePath(audioNode, 0),
                    ["audio_vae"] = g.CurrentAudioVae.Path,
                    ["identity_guidance_scale"] = 3, // TODO: Param?
                    ["start_percent"] = 0,
                    ["end_percent"] = 1
                });
                g.CurrentModel = g.CurrentModel.WithPath([refNode, 0]);
                g.FinalPrompt = [refNode, 1];
                g.FinalNegativePrompt = [refNode, 2];
            }
        }, -6);
        #region SAM2 Masking
        AddStep(g =>
        {
            if (!g.UserInput.TryGet(ComfyUIBackendExtension.Sam2PointCoordsPositive, out string coords) || string.IsNullOrWhiteSpace(coords) || coords == "[]")
            {
                return;
            }
            string negCoords = null;
            if (g.UserInput.TryGet(ComfyUIBackendExtension.Sam2PointCoordsNegative, out string negCoordsRaw) && !string.IsNullOrWhiteSpace(negCoordsRaw) && negCoordsRaw != "[]")
            {
                negCoords = negCoordsRaw;
            }
            JArray imageNodeActual = g.BasicInputImage?.Path;
            if (imageNodeActual is null)
            {
                return;
            }
            string modelNode = g.CreateNode("DownloadAndLoadSAM2Model", ComfyUIBackendExtension.Sam2ModelInputs());
            JObject segInputs = new()
            {
                ["sam2_model"] = NodePath(modelNode, 0),
                ["image"] = imageNodeActual,
                ["keep_model_loaded"] = true,
                ["coordinates_positive"] = coords
            };
            if (negCoords is not null)
            {
                segInputs["coordinates_negative"] = negCoords;
            }
            string segNode = g.CreateNode("Sam2Segmentation", segInputs);
            string postNode = g.CreateNode("SwarmSam2MaskPostProcess", new JObject()
            {
                ["mask"] = NodePath(segNode, 0),
                ["fill_holes"] = true,
                ["hole_kernel_size"] = 9
            });
            string maskNode = g.CreateNode("MaskToImage", new JObject()
            {
                ["mask"] = NodePath(postNode, 0)
            });
            new WGNodeData([maskNode, 0], g, WGNodeData.DT_IMAGE, g.CurrentCompat()).SaveOutput(null, null, "9");
            g.SkipFurtherSteps = true;
        }, -5.8);
        AddStep(g =>
        {
            if (!g.UserInput.TryGet(ComfyUIBackendExtension.Sam2BBox, out string bboxJson) || string.IsNullOrWhiteSpace(bboxJson))
            {
                return;
            }
            JArray imageNodeActual = g.BasicInputImage?.Path;
            if (imageNodeActual is null)
            {
                return;
            }
            string modelNode = g.CreateNode("DownloadAndLoadSAM2Model", ComfyUIBackendExtension.Sam2ModelInputs());
            string bboxNode = g.CreateNode("SwarmSam2BBoxFromJson", new JObject()
            {
                ["bbox_json"] = bboxJson
            });
            JObject segInputs = new()
            {
                ["sam2_model"] = NodePath(modelNode, 0),
                ["image"] = imageNodeActual,
                ["keep_model_loaded"] = true,
                ["bboxes"] = NodePath(bboxNode, 0)
            };
            string segNode = g.CreateNode("Sam2Segmentation", segInputs);
            string postNode = g.CreateNode("SwarmSam2MaskPostProcess", new JObject()
            {
                ["mask"] = NodePath(segNode, 0),
                ["fill_holes"] = true,
                ["hole_kernel_size"] = 5
            });
            string maskNode = g.CreateNode("MaskToImage", new JObject()
            {
                ["mask"] = NodePath(postNode, 0)
            });
            new WGNodeData([maskNode, 0], g, WGNodeData.DT_IMAGE, g.CurrentCompat()).SaveOutput(null, null, "9");
            g.SkipFurtherSteps = true;
        }, -5.7);
        #endregion
        #endregion
        #region Sampler
        AddStep(g =>
        {
            int steps = g.UserInput.Get(T2IParamTypes.Steps);
            bool noSkip = false;
            if (steps < 0)
            {
                noSkip = true;
                steps = 0;
            }
            int startStep = 0;
            int endStep = 10000;
            if (g.UserInput.TryGet(T2IParamTypes.InitImage, out Image _) && g.UserInput.TryGet(T2IParamTypes.InitImageCreativity, out double creativity))
            {
                startStep = (int)Math.Round(steps * (1 - creativity));
            }
            if (g.UserInput.TryGet(T2IParamTypes.RefinerMethod, out string method) && method == "StepSwap" && g.UserInput.TryGet(T2IParamTypes.RefinerControl, out double refinerControl))
            {
                endStep = (int)Math.Round(steps * (1 - refinerControl));
            }
            if (g.UserInput.TryGet(T2IParamTypes.EndStepsEarly, out double endEarly))
            {
                endStep = (int)(steps * (1 - endEarly));
            }
            double cfg = g.UserInput.Get(T2IParamTypes.CFGScale);
            if (!noSkip && (steps == 0 || endStep <= startStep))
            {
                g.CreateNode("SwarmJustLoadTheModelPlease", new JObject()
                {
                    ["model"] = g.CurrentModel.Path,
                    ["clip"] = g.CurrentTextEnc.Path,
                    ["vae"] = g.CurrentVae.Path
                });
            }
            else if (steps > 0 && Math.Min(endStep, steps) > startStep)
            {
                g.CurrentMedia = g.CurrentMedia.AsSamplingLatent(g.CurrentVae, g.CurrentAudioVae);
                g.CreateKSampler(g.CurrentModel.Path, g.FinalPrompt, g.FinalNegativePrompt, g.CurrentMedia.Path, cfg, steps, startStep, endStep,
                    g.UserInput.Get(T2IParamTypes.Seed), g.UserInput.Get(T2IParamTypes.RefinerMethod, "none") == "StepSwapNoisy", g.MainSamplerAddNoise, id: "10", isFirstSampler: true, sectionId: T2IParamInput.SectionID_BaseOnly);
                g.CurrentMedia = g.CurrentMedia.WithPath(["10", 0]);
                if (g.UserInput.Get(T2IParamTypes.UseReferenceOnly, false))
                {
                    string fromBatch = g.CreateNode("LatentFromBatch", new JObject()
                    {
                        ["samples"] = NodePath("10", 0),
                        ["batch_index"] = 1,
                        ["length"] = 1
                    });
                    g.CurrentMedia = g.CurrentMedia.WithPath([fromBatch, 0], WGNodeData.DT_LATENT_IMAGE);
                    g.CurrentMedia.Frames = null;
                }
            }
            if (g.FinalTrimLatent is not null)
            {
                g.CurrentMedia = g.CurrentMedia.AsLatentImage(g.CurrentVae);
                string trimmed = g.CreateNode("TrimVideoLatent", new JObject()
                {
                    ["samples"] = g.CurrentMedia.Path,
                    ["trim_amount"] = g.FinalTrimLatent
                });
                // TODO: Frame count update
                g.CurrentMedia = g.CurrentMedia.WithPath([trimmed, 0]);
                g.FinalTrimLatent = null;
            }
        }, -5);
        JArray doMaskShrinkApply(WorkflowGenerator g, JArray imgIn)
        {
            (string boundsNode, string croppedMask, string masked, string scaledImage) = g.MaskShrunkInfo;
            g.MaskShrunkInfo = new(null, null, null, null);
            if (boundsNode is not null)
            {
                imgIn = g.RecompositeCropped(boundsNode, [croppedMask, 0], g.BasicInputImage.Path, imgIn);
            }
            else if (g.UserInput.Get(T2IParamTypes.InitImageRecompositeMask, true) && g.FinalMask is not null && !g.NodeHelpers.ContainsKey("recomposite_mask_result"))
            {
                imgIn = g.CompositeMask(g.BasicInputImage.Path, imgIn, g.FinalMask);
            }
            g.NodeHelpers["recomposite_mask_result"] = $"{imgIn[0]}";
            return imgIn;
        }
        #endregion
        #region Refiner
        AddStep(g =>
        {
            if (g.UserInput.TryGet(T2IParamTypes.RefinerMethod, out string method)
                && g.UserInput.TryGet(T2IParamTypes.RefinerControl, out double refinerControl))
            {
                g.IsRefinerStage = true;
                WGNodeData origVae = g.CurrentVae;
                JArray prompt = g.FinalPrompt, negPrompt = g.FinalNegativePrompt;
                bool modelMustReencode = false;
                T2IModel baseModel = g.UserInput.Get(T2IParamTypes.Model);
                T2IModel refineModel = baseModel;
                string loaderNodeId = null;
                if (g.UserInput.TryGet(T2IParamTypes.RefinerModel, out T2IModel altRefineModel) && altRefineModel is not null)
                {
                    refineModel = altRefineModel;
                    modelMustReencode = true;
                    if (refineModel.ModelClass?.CompatClass == baseModel.ModelClass?.CompatClass)
                    {
                        modelMustReencode = false;
                    }
                    if (refineModel.ModelClass?.CompatClass?.ID == "stable-diffusion-xl-v1-refiner" && baseModel.ModelClass?.CompatClass?.ID == "stable-diffusion-xl-v1")
                    {
                        modelMustReencode = false;
                    }
                    loaderNodeId = "20";
                }
                if (g.UserInput.TryGet(T2IParamTypes.RefinerVAE, out _))
                {
                    modelMustReencode = true;
                }
                g.NoVAEOverride = refineModel.ModelClass?.CompatClass != baseModel.ModelClass?.CompatClass;
                g.FinalLoadedModel = refineModel;
                g.FinalLoadedModelList = [refineModel];
                (g.FinalLoadedModel, g.CurrentModel, g.CurrentTextEnc, g.CurrentVae) = g.CreateModelLoader(refineModel, "Refiner", loaderNodeId, sectionId: T2IParamInput.SectionID_Refiner);
                g.NoVAEOverride = false;
                prompt = g.CreateConditioning(g.UserInput.Get(T2IParamTypes.Prompt), g.CurrentTextEnc.Path, g.FinalLoadedModel, true, isRefiner: true);
                negPrompt = g.CreateConditioning(g.UserInput.Get(T2IParamTypes.NegativePrompt), g.CurrentTextEnc.Path, g.FinalLoadedModel, false, isRefiner: true);
                bool doSave = g.UserInput.Get(T2IParamTypes.OutputIntermediateImages, false);
                bool doUspcale = g.UserInput.TryGet(T2IParamTypes.RefinerUpscale, out double refineUpscale) && refineUpscale != 1;
                string upscaleMethod = g.UserInput.Get(ComfyUIBackendExtension.RefinerUpscaleMethod, "None");
                // TODO: Better same-VAE check
                bool doPixelUpscale = doUspcale && (upscaleMethod.StartsWith("pixel-") || upscaleMethod.StartsWith("model-"));
                int width = (int)Math.Round(g.UserInput.GetImageWidth() * refineUpscale);
                int height = (int)Math.Round(g.UserInput.GetImageHeight() * refineUpscale);
                width = (width / 16) * 16; // avoid unworkable output sizes
                height = (height / 16) * 16;
                if (modelMustReencode || doPixelUpscale || doSave || g.MaskShrunkInfo.BoundsNode is not null)
                {
                    WGNodeData decoded = g.CurrentMedia.DecodeLatents(origVae, false, "24");
                    JArray maskShrunk = doMaskShrinkApply(g, decoded.Path);
                    decoded = decoded.WithPath(maskShrunk);
                    g.CurrentMedia = decoded;
                    if (doSave)
                    {
                        decoded.SaveOutput(null, null, id: "29");
                    }
                    if (doPixelUpscale)
                    {
                        if (upscaleMethod.StartsWith("pixel-"))
                        {
                            g.CreateNode("ImageScale", new JObject()
                            {
                                ["image"] = decoded.Path,
                                ["width"] = width,
                                ["height"] = height,
                                ["upscale_method"] = upscaleMethod.After("pixel-"),
                                ["crop"] = "disabled"
                            }, "26");
                        }
                        else
                        {
                            g.CreateNode("UpscaleModelLoader", new JObject()
                            {
                                ["model_name"] = upscaleMethod.After("model-")
                            }, "27");
                            g.CreateNode("ImageUpscaleWithModel", new JObject()
                            {
                                ["upscale_model"] = NodePath("27", 0),
                                ["image"] = decoded.Path
                            }, "28");
                            g.CreateNode("ImageScale", new JObject()
                            {
                                ["image"] = NodePath("28", 0),
                                ["width"] = width,
                                ["height"] = height,
                                ["upscale_method"] = "lanczos",
                                ["crop"] = "disabled"
                            }, "26");
                        }
                        decoded = decoded.WithPath(["26", 0]);
                        decoded.Width = width;
                        decoded.Height = height;
                        if (refinerControl <= 0)
                        {
                            g.CurrentMedia = decoded;
                            return;
                        }
                    }
                    if (modelMustReencode || doPixelUpscale)
                    {
                        g.CurrentMedia = decoded.EncodeToLatent(g.CurrentVae, "25");
                    }
                }
                if (doUspcale && upscaleMethod.StartsWith("latent-"))
                {
                    g.CurrentMedia = g.CurrentMedia.AsLatentImage(g.CurrentVae);
                    g.CreateNode("LatentUpscaleBy", new JObject()
                    {
                        ["samples"] = g.CurrentMedia.Path,
                        ["upscale_method"] = upscaleMethod.After("latent-"),
                        ["scale_by"] = refineUpscale
                    }, "26");
                    g.CurrentMedia = g.CurrentMedia.WithPath(["26", 0]);
                    g.CurrentMedia.Width = width;
                    g.CurrentMedia.Height = height;
                }
                else if (doUspcale && upscaleMethod.StartsWith("latentmodel-"))
                {
                    g.CreateNode("LatentUpscaleModelLoader", new JObject()
                    {
                        ["model_name"] = upscaleMethod.After("latentmodel-")
                    }, "27");
                    if (g.IsHunyuanVideo15())
                    {
                        g.CreateNode("HunyuanVideo15LatentUpscaleWithModel", new JObject()
                        {
                            ["model"] = NodePath("27", 0),
                            ["samples"] = g.CurrentMedia.Path,
                            ["upscale_method"] = "bilinear",
                            ["width"] = width,
                            ["height"] = height,
                            ["crop"] = "disabled"
                        }, "26");
                        g.CurrentMedia = g.CurrentMedia.WithPath(["26", 0], WGNodeData.DT_LATENT_VIDEO);
                    }
                    else if (g.IsLTXV2())
                    {
                        g.CurrentMedia = g.CurrentMedia.AsLatentImage(g.CurrentVae);
                        string cropGuides = g.CreateNode("LTXVCropGuides", new JObject()
                        {
                            ["positive"] = prompt,
                            ["negative"] = negPrompt,
                            ["latent"] = g.CurrentMedia.Path
                        });
                        prompt = [cropGuides, 0];
                        negPrompt = [cropGuides, 1];
                        g.CreateNode("LTXVLatentUpsampler", new JObject()
                        {
                            ["vae"] = g.CurrentVae.Path,
                            ["samples"] = NodePath(cropGuides, 2),
                            ["upscale_model"] = NodePath("27", 0)
                        }, "26");
                        g.CurrentMedia = g.CurrentMedia.WithPath(["26", 0], WGNodeData.DT_LATENT_VIDEO);
                    }
                    else
                    {
                        throw new SwarmUserErrorException($"Cannot latent-upscale with a model for {g.CurrentCompatClass()}, check your Refiner Upscale Method parameter");
                    }
                    g.CurrentMedia.Width = width;
                    g.CurrentMedia.Height = height;
                }
                WGNodeData model = g.CurrentModel;
                if (g.UserInput.TryGet(ComfyUIBackendExtension.RefinerHyperTile, out int tileSize))
                {
                    string hyperTileNode = g.CreateNode("HyperTile", new JObject()
                    {
                        ["model"] = model.Path,
                        ["tile_size"] = tileSize,
                        ["swap_size"] = 2, // TODO: Do these other params matter?
                        ["max_depth"] = 0,
                        ["scale_depth"] = false
                    });
                    model = model.WithPath([hyperTileNode, 0]);
                }
                g.CurrentMedia = g.CurrentMedia.AsSamplingLatent(g.CurrentVae, g.CurrentAudioVae);
                int steps = g.UserInput.Get(T2IParamTypes.RefinerSteps, g.UserInput.Get(T2IParamTypes.Steps, 20, sectionId: T2IParamInput.SectionID_Refiner), sectionId: T2IParamInput.SectionID_Refiner);
                double cfg = g.UserInput.Get(T2IParamTypes.RefinerCFGScale, g.UserInput.Get(T2IParamTypes.CFGScale, 7, sectionId: T2IParamInput.SectionID_Refiner), sectionId: T2IParamInput.SectionID_Refiner);
                string explicitSampler = g.UserInput.Get(ComfyUIBackendExtension.SamplerParam, null, sectionId: T2IParamInput.SectionID_Refiner, includeBase: false) ?? g.UserInput.Get(ComfyUIBackendExtension.RefinerSamplerParam, null);
                string explicitScheduler = g.UserInput.Get(ComfyUIBackendExtension.SchedulerParam, null, sectionId: T2IParamInput.SectionID_Refiner, includeBase: false) ?? g.UserInput.Get(ComfyUIBackendExtension.RefinerSchedulerParam, null);
                g.CreateKSampler(model.Path, prompt, negPrompt, g.CurrentMedia.Path, cfg, steps, (int)Math.Round(steps * (1 - refinerControl)), 10000,
                    g.UserInput.Get(T2IParamTypes.Seed) + 1, false, method != "StepSwapNoisy", id: "23", doTiled: g.UserInput.Get(T2IParamTypes.RefinerDoTiling, false),
                    explicitSampler: explicitSampler, explicitScheduler: explicitScheduler, sectionId: T2IParamInput.SectionID_Refiner);
                g.CurrentMedia = g.CurrentMedia.WithPath(["23", 0]);
                g.IsRefinerStage = false;
            }
        }, -4);
        #endregion
        #region VAEDecode
        AddStep(g =>
        {
            g.CurrentMedia = g.CurrentMedia.DecodeLatents(g.CurrentVae, null, "8");
            JArray maskShrinkApply = doMaskShrinkApply(g, g.CurrentMedia.Path);
            g.CurrentMedia = g.CurrentMedia.WithPath(maskShrinkApply);
        }, 1);
        #endregion
        #region Segmentation Processing
        void RunSegmentationProcessing(WorkflowGenerator g, bool isBeforeRefiner)
        {
            PromptRegion.Part[] parts = [.. new PromptRegion(g.UserInput.Get(T2IParamTypes.Prompt, "")).Parts.Where(p => p.Type == PromptRegion.PartType.Segment)];
            if (parts.Any())
            {
                g.CurrentMedia = g.CurrentMedia.AsRawImage(g.CurrentVae);
                if (g.UserInput.Get(T2IParamTypes.OutputIntermediateImages, false))
                {
                    g.CurrentMedia.SaveOutput(g.CurrentVae, g.CurrentAudioVae, id: g.GetStableDynamicID(50000, 0));
                }
                T2IModel t2iModel = g.FinalLoadedModel;
                WGNodeData model = g.CurrentModel, clip = g.CurrentTextEnc, vae = g.CurrentVae;
                if (g.UserInput.TryGet(T2IParamTypes.SegmentModel, out T2IModel segmentModel))
                {
                    if (segmentModel.ModelClass?.CompatClass != t2iModel.ModelClass?.CompatClass)
                    {
                        g.NoVAEOverride = true;
                    }
                    t2iModel = segmentModel;
                    g.FinalLoadedModel = segmentModel;
                    (t2iModel, model, clip, vae) = g.CreateModelLoader(t2iModel, "Refiner", sectionId: parts[0].ContextID);
                    g.FinalLoadedModel = t2iModel;
                    g.CurrentModel = model;
                }
                PromptRegion negativeRegion = new(g.UserInput.Get(T2IParamTypes.NegativePrompt, ""));
                PromptRegion.Part[] negativeParts = [.. negativeRegion.Parts.Where(p => p.Type == PromptRegion.PartType.Segment)];
                for (int i = 0; i < parts.Length; i++)
                {
                    PromptRegion.Part part = parts[i];
                    string[] segmentSections = part.DataText.Split('|', StringSplitOptions.RemoveEmptyEntries | StringSplitOptions.TrimEntries);
                    string segmentNode = null;
                    if (segmentSections.Length == 0)
                    {
                        segmentSections = [""];
                    }
                    foreach (string dataText in segmentSections)
                    {
                        string newSegmentNode = null;
                        if (dataText.StartsWith("yolo-"))
                        {
                            string fullname = dataText.After("yolo-");
                            string[] modelParts = fullname.Split(':');
                            fullname = modelParts[0];
                            string classFilter = modelParts.Length > 1 ? modelParts[1] : "";
                            (string mname, string indexText) = fullname.BeforeAndAfterLast('-');
                            if (!string.IsNullOrWhiteSpace(indexText) && int.TryParse(indexText, out int index))
                            {
                                fullname = mname;
                            }
                            else
                            {
                                index = 0;
                            }
                            if (part.Strength > 0.999)
                            {
                                Logs.Warning($"Yolo confidence threshold is set to 1. This was recommended syntax before yolo thresholds were supported, but is no longer valid. Swarm will automatically reset the value to default (0.25) instead.");
                                part.Strength = 0.25;
                            }
                            newSegmentNode = g.CreateNode("SwarmYoloDetection", new JObject()
                            {
                                ["image"] = g.CurrentMedia.Path,
                                ["model_name"] = fullname,
                                ["index"] = index,
                                ["class_filter"] = classFilter,
                                ["sort_order"] = g.UserInput.Get(T2IParamTypes.SegmentSortOrder, "left-right"),
                                ["threshold"] = Math.Abs(part.Strength)
                            });
                        }
                        else
                        {
                            newSegmentNode = g.CreateNode("SwarmClipSeg", new JObject()
                            {
                                ["images"] = g.CurrentMedia.Path,
                                ["match_text"] = dataText,
                                ["threshold"] = Math.Abs(part.Strength)
                            });
                        }
                        if (segmentSections.Length > 1 && g.UserInput.Get(T2IParamTypes.SaveSegmentMask, false))
                        {
                            string imageNode = g.CreateNode("MaskToImage", new JObject()
                            {
                                ["mask"] = NodePath(newSegmentNode, 0)
                            });
                            new WGNodeData([imageNode, 0], g, WGNodeData.DT_IMAGE, g.CurrentCompat()).SaveOutput(null, null, g.GetStableDynamicID(50000, 0));
                        }
                        if (segmentNode is null)
                        {
                            segmentNode = newSegmentNode;
                        }
                        else
                        {
                            segmentNode = g.CreateNode("MaskComposite", new JObject()
                            {
                                ["destination"] = NodePath(segmentNode, 0),
                                ["source"] = NodePath(newSegmentNode, 0),
                                ["operation"] = "add",
                                ["x"] = 0,
                                ["y"] = 0
                            });
                        }
                    }
                    if (part.Strength < 0)
                    {
                        segmentNode = g.CreateNode("InvertMask", new JObject()
                        {
                            ["mask"] = NodePath(segmentNode, 0)
                        });
                    }
                    int blurAmt = g.UserInput.Get(T2IParamTypes.SegmentMaskBlur, 10);
                    if (blurAmt > 0)
                    {
                        segmentNode = g.CreateNode("SwarmMaskBlur", new JObject()
                        {
                            ["mask"] = NodePath(segmentNode, 0),
                            ["blur_radius"] = blurAmt,
                            ["sigma"] = 1
                        });
                    }
                    int growAmt = g.UserInput.Get(T2IParamTypes.SegmentMaskGrow, 16);
                    if (growAmt > 0)
                    {
                        segmentNode = g.CreateNode("GrowMask", new JObject()
                        {
                            ["mask"] = NodePath(segmentNode, 0),
                            ["expand"] = growAmt,
                            ["tapered_corners"] = true
                        });
                    }
                    if (g.UserInput.Get(T2IParamTypes.SaveSegmentMask, false))
                    {
                        string imageNode = g.CreateNode("MaskToImage", new JObject()
                        {
                            ["mask"] = NodePath(segmentNode, 0)
                        });
                        new WGNodeData([imageNode, 0], g, WGNodeData.DT_IMAGE, g.CurrentCompat()).SaveOutput(null, null, g.GetStableDynamicID(50000, 0));
                    }
                    int oversize = g.UserInput.Get(T2IParamTypes.SegmentMaskOversize, 16);
                    g.MaskShrunkInfo = g.CreateImageMaskCrop([segmentNode, 0], g.CurrentMedia.Path, oversize, vae.Path, g.FinalLoadedModel, thresholdMax: g.UserInput.Get(T2IParamTypes.SegmentThresholdMax, 1));
                    g.EnableDifferential();
                    if (part.ContextID > 0)
                    {
                        (JArray newModel, JArray newClip) = g.LoadLorasForConfinement(part.ContextID, g.CurrentModel.Path, clip.Path);
                        model = model.WithPath(newModel);
                        clip = clip.WithPath(newClip);
                    }
                    JArray prompt = g.CreateConditioning(part.Prompt, clip.Path, t2iModel, true);
                    string neg = negativeParts.FirstOrDefault(p => p.DataText == part.DataText)?.Prompt ?? negativeRegion.GlobalPrompt;
                    JArray negPrompt = g.CreateConditioning(neg, clip.Path, t2iModel, false);
                    int steps = g.UserInput.GetNullable(T2IParamTypes.Steps, part.ContextID, false) ?? g.UserInput.GetNullable(T2IParamTypes.SegmentSteps, part.ContextID) ?? g.UserInput.GetNullable(T2IParamTypes.RefinerSteps, part.ContextID) ?? g.UserInput.Get(T2IParamTypes.Steps, 20, sectionId: part.ContextID);
                    int startStep = (int)Math.Round(steps * (1 - part.Strength2));
                    long seed = g.UserInput.Get(T2IParamTypes.Seed) + 2 + i;
                    double cfg = g.UserInput.GetNullable(T2IParamTypes.CFGScale, part.ContextID, false) ?? g.UserInput.GetNullable(T2IParamTypes.SegmentCFGScale, part.ContextID) ?? g.UserInput.GetNullable(T2IParamTypes.RefinerCFGScale, part.ContextID) ?? g.UserInput.Get(T2IParamTypes.CFGScale, 7, sectionId: part.ContextID);
                    WGNodeData beforeImage = g.CurrentMedia;
                    string sampler = g.CreateKSampler(model.Path, prompt, negPrompt, [g.MaskShrunkInfo.MaskedLatent, 0], cfg, steps, startStep, 10000, seed, false, true, sectionId: part.ContextID);
                    g.CurrentMedia = g.CurrentMedia.WithPath([sampler, 0], WGNodeData.DT_LATENT_IMAGE);
                    g.CurrentMedia = g.CurrentMedia.AsRawImage(vae);
                    JArray composited = g.RecompositeCropped(g.MaskShrunkInfo.BoundsNode, [g.MaskShrunkInfo.CroppedMask, 0], beforeImage.Path, g.CurrentMedia.Path);
                    g.CurrentMedia = g.CurrentMedia.WithPath(composited);
                    g.MaskShrunkInfo = new(null, null, null, null);
                }
            }
        }
        AddStep(g =>
        {
            if (g.UserInput.Get(T2IParamTypes.SegmentApplyAfter, "Refiner") != "Base")
            {
                return;
            }
            PromptRegion.Part[] parts = [.. new PromptRegion(g.UserInput.Get(T2IParamTypes.Prompt, "")).Parts.Where(p => p.Type == PromptRegion.PartType.Segment)];
            if (!parts.Any())
            {
                return;
            }
            RunSegmentationProcessing(g, isBeforeRefiner: true);
        }, -4.5);
        AddStep(g =>
        {
            if (g.UserInput.Get(T2IParamTypes.SegmentApplyAfter, "Refiner") != "Refiner")
            {
                return;
            }
            RunSegmentationProcessing(g, isBeforeRefiner: false);
        }, 5);
        #endregion
        #region SaveImage
        AddStep(g =>
        {
            PromptRegion.Part[] parts = [.. new PromptRegion(g.UserInput.Get(T2IParamTypes.Prompt, "")).Parts.Where(p => p.Type == PromptRegion.PartType.ClearSegment)];
            g.CurrentMedia = g.CurrentMedia.DecodeLatents(g.CurrentVae, null);
            foreach (PromptRegion.Part part in parts)
            {
                if (g.UserInput.Get(T2IParamTypes.OutputIntermediateImages, false))
                {
                    g.CurrentMedia.SaveOutput(g.CurrentVae, g.CurrentAudioVae, g.GetStableDynamicID(50000, 0));
                }
                string segmentNode = g.CreateNode("SwarmClipSeg", new JObject()
                {
                    ["images"] = g.CurrentMedia.Path,
                    ["match_text"] = part.DataText,
                    ["threshold"] = Math.Abs(part.Strength)
                });
                if (part.Strength < 0)
                {
                    segmentNode = g.CreateNode("InvertMask", new JObject()
                    {
                        ["mask"] = NodePath(segmentNode, 0)
                    });
                }
                string blurNode = g.CreateNode("SwarmMaskBlur", new JObject()
                {
                    ["mask"] = NodePath(segmentNode, 0),
                    ["blur_radius"] = 10,
                    ["sigma"] = 1
                });
                string thresholded = g.CreateNode("SwarmMaskThreshold", new JObject()
                {
                    ["mask"] = NodePath(blurNode, 0),
                    ["min"] = 0.2,
                    ["max"] = 0.6
                });
                string joined = g.CreateNode("JoinImageWithAlpha", new JObject()
                {
                    ["image"] = g.CurrentMedia.Path,
                    ["alpha"] = NodePath(thresholded, 0)
                });
                g.CurrentMedia = g.CurrentMedia.WithPath([joined, 0]);
            }
            if (g.UserInput.Get(T2IParamTypes.RemoveBackground, false))
            {
                if (g.UserInput.Get(T2IParamTypes.OutputIntermediateImages, false))
                {
                    g.CurrentMedia.SaveOutput(g.CurrentVae, g.CurrentAudioVae, id: g.GetStableDynamicID(50000, 0));
                }
                string removed = g.CreateNode("SwarmRemBg", new JObject()
                {
                    ["images"] = g.CurrentMedia.Path
                });
                g.CurrentMedia = g.CurrentMedia.WithPath([removed, 0]);
            }
            if (g.UserInput.SourceSession is null && g.UserInput.Get(T2IParamTypes.DoNotSave, false) && g.UserInput.Get(T2IParamTypes.Steps) == 0 && !g.UserInput.TryGet(T2IParamTypes.RefinerModel, out _))
            {
                // We don't actually want an image we're just aggressively loading a model or something
            }
            else
            {
                bool willHaveFollowupVideo = g.UserInput.TryGet(T2IParamTypes.VideoModel, out _) || g.UserInput.Get(T2IParamTypes.Prompt, "").Contains("<extend:");
                // Heuristic check for if this is an Init Image with no further processing, ie the initial image save is redundant because we're just wanting to extend a presaved image to a video
                bool formedFromSingleImage = g.UserInput.Get(T2IParamTypes.InitImageCreativity, -1) == 0 && !g.UserInput.Get(T2IParamTypes.OutputIntermediateImages, false) && !g.UserInput.TryGet(T2IParamTypes.RefinerMethod, out _);
                if (g.IsVideoModel() && !formedFromSingleImage && !willHaveFollowupVideo)
                {
                    if (g.UserInput.TryGet(T2IParamTypes.TrimVideoStartFrames, out _) || g.UserInput.TryGet(T2IParamTypes.TrimVideoEndFrames, out _))
                    {
                        string trimNode = g.CreateNode("SwarmTrimFrames", new JObject()
                        {
                            ["image"] = g.CurrentMedia.Path,
                            ["trim_start"] = g.UserInput.Get(T2IParamTypes.TrimVideoStartFrames, 0),
                            ["trim_end"] = g.UserInput.Get(T2IParamTypes.TrimVideoEndFrames, 0)
                        });
                        g.CurrentMedia = g.CurrentMedia.WithPath([trimNode, 0]);
                    }
                    if (g.UserInput.TryGet(ComfyUIBackendExtension.VideoFrameInterpolationMethod, out string method)
                        && g.UserInput.TryGet(ComfyUIBackendExtension.VideoFrameInterpolationMultiplier, out int mult) && mult > 1
                        && g.UserInput.Get(T2IParamTypes.Text2VideoFrames, 99) > 1)
                    {
                        if (g.UserInput.Get(T2IParamTypes.OutputIntermediateImages, false))
                        {
                            g.CurrentMedia.SaveOutput(g.CurrentVae, g.CurrentAudioVae, id: g.GetStableDynamicID(50000, 0));
                        }
                        JArray newInterp = g.DoInterpolation(g.CurrentMedia.Path, method, mult);
                        g.CurrentMedia = g.CurrentMedia.WithPath(newInterp);
                        int fps = g.CurrentMedia.GetRawFPS() ?? g.Text2VideoFPS();
                        fps *= mult;
                        g.CurrentMedia.FPS = fps;
                        g.T2VFPSOverride = fps;
                    }
                }
                string nodeId = "9";
                if (willHaveFollowupVideo)
                {
                    nodeId = "30";
                    if (formedFromSingleImage)
                    {
                        nodeId = null;
                    }
                }
                if (nodeId is not null)
                {
                    g.CurrentMedia.SaveOutput(g.CurrentVae, g.CurrentAudioVae, nodeId);
                }
            }
        }, 10);
        #endregion
        #region Image To Video
        AddStep(g =>
        {
            if (g.UserInput.TryGet(T2IParamTypes.VideoModel, out T2IModel vidModel))
            {
                int? frames = g.UserInput.TryGet(T2IParamTypes.VideoFrames, out int framesRaw) ? framesRaw : null;
                int? videoFps = g.UserInput.TryGet(T2IParamTypes.VideoFPS, out int fpsRaw) ? fpsRaw : null;
                double? videoCfg = g.UserInput.GetNullable(T2IParamTypes.CFGScale, T2IParamInput.SectionID_Video, false) ?? g.UserInput.GetNullable(T2IParamTypes.VideoCFG, T2IParamInput.SectionID_Video);
                int steps = g.UserInput.GetNullable(T2IParamTypes.Steps, T2IParamInput.SectionID_Video, false) ?? g.UserInput.Get(T2IParamTypes.VideoSteps, 20, sectionId: T2IParamInput.SectionID_Video);
                string resFormat = g.UserInput.Get(T2IParamTypes.VideoResolution, "Model Preferred");
                long seed = g.UserInput.Get(T2IParamTypes.Seed) + 42;
                string prompt = g.UserInput.Get(T2IParamTypes.Prompt, "");
                string negPrompt = g.UserInput.Get(T2IParamTypes.NegativePrompt, "");
                int batchInd = -1, batchLen = -1;
                if (g.UserInput.TryGet(T2IParamTypes.Video2VideoCreativity, out _))
                {
                    batchInd = 0;
                    batchLen = 1;
                }
                int width = vidModel.StandardWidth <= 0 ? 1024 : vidModel.StandardWidth;
                int height = vidModel.StandardHeight <= 0 ? 576 : vidModel.StandardHeight;
                int imageWidth = g.UserInput.GetImageWidth();
                int imageHeight = g.UserInput.GetImageHeight();
                int resPrecision = 64;
                if (vidModel.ModelClass?.CompatClass?.ID == "hunyuan-video")
                {
                    resPrecision = 16; // wants 720x720, which is wonky x16 and not x32 or x64
                }
                if (resFormat == "Image Aspect, Model Res")
                {
                    if (width == 1024 && height == 576 && imageWidth == 1344 && imageHeight == 768)
                    {
                        width = 1024;
                        height = 576;
                    }
                    else
                    {
                        (width, height) = Utilities.ResToModelFit(imageWidth, imageHeight, width * height, resPrecision);
                    }
                }
                else if (resFormat == "Image")
                {
                    width = imageWidth;
                    height = imageHeight;
                    if (g.UserInput.TryGet(T2IParamTypes.RefinerUpscale, out double scale))
                    {
                        width = (int)Math.Round(width * scale);
                        height = (int)Math.Round(height * scale);
                    }
                }
                g.CurrentMedia = g.CurrentMedia.AsRawImage(g.CurrentVae);
                ImageToVideoGenInfo genInfo = new()
                {
                    Generator = g,
                    VideoModel = vidModel,
                    VideoSwapModel = g.UserInput.Get(T2IParamTypes.VideoSwapModel, null),
                    VideoSwapPercent = g.UserInput.Get(T2IParamTypes.VideoSwapPercent, 0.5),
                    Frames = frames,
                    VideoCFG = videoCfg,
                    VideoFPS = videoFps,
                    Width = width,
                    Height = height,
                    Prompt = prompt,
                    NegativePrompt = negPrompt,
                    Steps = steps,
                    Seed = seed,
                    BatchIndex = batchInd,
                    BatchLen = batchLen,
                    ContextID = T2IParamInput.SectionID_Video,
                    VideoEndFrame = g.UserInput.Get(T2IParamTypes.VideoEndFrame, null)
                };
                if (g.UserInput.TryGet(T2IParamTypes.Video2VideoCreativity, out double v2vCreativity))
                {
                    genInfo.StartStep = (int)Math.Floor(steps * (1 - v2vCreativity));
                }
                g.CreateImageToVideo(genInfo);
                g.CurrentMedia = g.CurrentMedia.AsRawImage(genInfo.Vae);
                bool hasExtend = prompt.Contains("<extend:");
                if (!hasExtend && g.UserInput.TryGet(ComfyUIBackendExtension.VideoFrameInterpolationMethod, out string method) && g.UserInput.TryGet(ComfyUIBackendExtension.VideoFrameInterpolationMultiplier, out int mult) && mult > 1)
                {
                    if (g.UserInput.Get(T2IParamTypes.OutputIntermediateImages, false))
                    {
                        g.CurrentMedia.SaveOutput(genInfo.Vae, g.CurrentAudioVae, g.GetStableDynamicID(50000, 0));
                    }
                    JArray interpolated = g.DoInterpolation(g.CurrentMedia.Path, method, mult);
                    g.CurrentMedia = g.CurrentMedia.WithPath(interpolated);
                    g.CurrentMedia.FPS = genInfo.VideoFPS.HasValue ? genInfo.VideoFPS.Value * mult : null;
                }
                string nodeId = "9";
                if (hasExtend)
                {
                    nodeId = $"{g.GetStableDynamicID(50000, 0)}";
                }
                g.CurrentMedia.SaveOutput(genInfo.Vae, g.CurrentAudioVae, nodeId);
            }
        }, 11);
        #endregion
        #region Extend Video
        AddStep(g =>
        {
            WGNodeData ensureAttachedAudio(WGNodeData media)
            {
                if (media?.AttachedAudio is null || media.AttachedAudio.DataType == WGNodeData.DT_AUDIO || g.CurrentAudioVae is null)
                {
                    return media;
                }
                WGNodeData dup = media.Duplicate();
                dup.AttachedAudio = media.AttachedAudio.DecodeLatents(g.CurrentAudioVae, true);
                return dup;
            }
            WGNodeData appendAudio(WGNodeData combinedAudio, WGNodeData nextAudio)
            {
                if (combinedAudio is null)
                {
                    return nextAudio;
                }
                if (nextAudio is null || combinedAudio.DataType != WGNodeData.DT_AUDIO || nextAudio.DataType != WGNodeData.DT_AUDIO)
                {
                    return combinedAudio;
                }
                string concatNode = g.CreateNode("AudioConcat", new JObject()
                {
                    ["audio1"] = combinedAudio.Path,
                    ["audio2"] = nextAudio.Path,
                    ["direction"] = "after"
                });
                return combinedAudio.WithPath([concatNode, 0], WGNodeData.DT_AUDIO, combinedAudio.Compat ?? nextAudio.Compat);
            }
            string fullRawPrompt = g.UserInput.Get(T2IParamTypes.Prompt, "");
            if (fullRawPrompt.Contains("<extend:"))
            {
                g.CurrentMedia = g.CurrentMedia.AsRawImage(g.CurrentVae);
                string negPrompt = g.UserInput.Get(T2IParamTypes.NegativePrompt, "");
                long seed = g.UserInput.Get(T2IParamTypes.Seed) + 600;
                int? videoFps = g.UserInput.TryGet(T2IParamTypes.VideoFPS, out int fpsRaw) ? fpsRaw : null;
                string format = g.UserInput.Get(T2IParamTypes.VideoExtendFormat, "mp4").ToLowerFast();
                int frameExtendOverlap = g.UserInput.Get(T2IParamTypes.VideoExtendFrameOverlap, 9);
                bool saveIntermediate = g.UserInput.Get(T2IParamTypes.OutputIntermediateImages, false);
                T2IModel extendModel = g.UserInput.Get(T2IParamTypes.VideoExtendModel, null) ?? throw new SwarmUserErrorException("You have an '<extend:' block in your prompt, but you don't have a 'Video Extend Model' selected.");
                PromptRegion regionalizer = new(fullRawPrompt);
                WGNodeData conjoinedLast = g.CurrentMedia;
                WGNodeData conjoinedAudio = ensureAttachedAudio(conjoinedLast)?.AttachedAudio;
                string getWidthNode = g.CreateNode("SwarmImageWidth", new JObject()
                {
                    ["image"] = g.CurrentMedia.Path
                });
                JArray width = [getWidthNode, 0];
                string getHeightNode = g.CreateNode("SwarmImageHeight", new JObject()
                {
                    ["image"] = g.CurrentMedia.Path
                });
                JArray height = [getHeightNode, 0];
                PromptRegion.Part[] parts = [.. regionalizer.Parts.Where(p => p.Type == PromptRegion.PartType.Extend)];
                for (int i = 0; i < parts.Length; i++)
                {
                    PromptRegion.Part part = parts[i];
                    double cfg = g.UserInput.GetNullable(T2IParamTypes.CFGScale, part.ContextID, false) ?? g.UserInput.GetNullable(T2IParamTypes.VideoCFG, part.ContextID) ?? g.UserInput.Get(T2IParamTypes.CFGScale, 7);
                    int steps = g.UserInput.GetNullable(T2IParamTypes.Steps, part.ContextID, false) ?? g.UserInput.GetNullable(T2IParamTypes.VideoSteps, part.ContextID) ?? g.UserInput.Get(T2IParamTypes.Steps, 20);
                    seed++;
                    int? frames = int.Parse(part.DataText);
                    string prompt = part.Prompt;
                    string frameCountNode = g.CreateNode("SwarmCountFrames", new JObject()
                    {
                        ["image"] = g.CurrentMedia.Path
                    });
                    JArray frameCount = [frameCountNode, 0];
                    string fromEndCountNode = g.CreateNode("SwarmIntAdd", new JObject()
                    {
                        ["a"] = frameCount,
                        ["b"] = -frameExtendOverlap
                    });
                    JArray fromEndCount = [fromEndCountNode, 0];
                    string partialBatchNode = g.CreateNode("ImageFromBatch", new JObject()
                    {
                        ["image"] = g.CurrentMedia.Path,
                        ["batch_index"] = fromEndCount,
                        ["length"] = frameExtendOverlap
                    });
                    JArray partialBatch = [partialBatchNode, 0];
                    g.CurrentMedia = g.CurrentMedia.WithPath(partialBatch);
                    ImageToVideoGenInfo genInfo = new()
                    {
                        Generator = g,
                        VideoModel = extendModel,
                        VideoSwapModel = g.UserInput.Get(T2IParamTypes.VideoExtendSwapModel, null),
                        VideoSwapPercent = g.UserInput.Get(T2IParamTypes.VideoExtendSwapPercent, 0.5),
                        Frames = frames,
                        VideoCFG = cfg,
                        VideoFPS = videoFps,
                        Width = width,
                        Height = height,
                        Prompt = prompt,
                        NegativePrompt = negPrompt,
                        Steps = steps,
                        Seed = seed,
                        BatchIndex = 0,
                        BatchLen = frameExtendOverlap,
                        ContextID = part.ContextID
                    };
                    g.CreateImageToVideo(genInfo);
                    g.CurrentMedia = g.CurrentMedia.AsRawImage(genInfo.Vae);
                    WGNodeData stageWithAudio = ensureAttachedAudio(g.CurrentMedia);
                    videoFps = genInfo.VideoFPS;
                    g.CurrentMedia.FPS = videoFps ?? g.CurrentMedia.FPS;
                    if (saveIntermediate)
                    {
                        g.CurrentMedia.SaveOutput(g.CurrentVae, g.CurrentAudioVae, g.GetStableDynamicID(50000, 0));
                    }
                    string cutNode = g.CreateNode("ImageFromBatch", new JObject()
                    {
                        ["image"] = g.CurrentMedia.Path,
                        ["batch_index"] = frameExtendOverlap,
                        ["length"] = frames.Value - frameExtendOverlap
                    });
                    g.CurrentMedia = g.CurrentMedia.WithPath([cutNode, 0]);
                    conjoinedAudio = appendAudio(conjoinedAudio, stageWithAudio?.AttachedAudio);
                    string batchedNode = g.CreateNode("ImageBatch", new JObject()
                    {
                        ["image1"] = conjoinedLast.Path,
                        ["image2"] = g.CurrentMedia.Path
                    });
                    conjoinedLast = conjoinedLast.WithPath([batchedNode, 0]);
                    conjoinedLast.AttachedAudio = conjoinedAudio;
                }
                g.CurrentMedia = conjoinedLast;
                g.CurrentMedia.FPS = videoFps ?? g.CurrentMedia.FPS;
                if (g.UserInput.TryGet(ComfyUIBackendExtension.VideoFrameInterpolationMethod, out string method) && g.UserInput.TryGet(ComfyUIBackendExtension.VideoFrameInterpolationMultiplier, out int mult) && mult > 1)
                {
                    if (saveIntermediate)
                    {
                        g.CurrentMedia.SaveOutput(g.CurrentVae, g.CurrentAudioVae, g.GetStableDynamicID(50000, 0));
                    }
                    JArray interp = g.DoInterpolation(g.CurrentMedia.Path, method, mult);
                    g.CurrentMedia = g.CurrentMedia.WithPath(interp);
                    videoFps *= mult;
                    g.CurrentMedia.FPS = videoFps;
                }
                g.CurrentMedia.SaveOutput(g.CurrentVae, g.CurrentAudioVae, "9");
            }
        }, 12);
        #endregion
        #region Post-Cleanup
        AddStep(g =>
        {
            g.RunOnNodesOfClass("SwarmKSampler", (id, data) =>
            {
                if (data["inputs"]["start_at_step"].Value<int>() >= data["inputs"]["steps"].Value<int>())
                {
                    g.ReplaceNodeConnection([id, 0], data["inputs"]["latent_image"] as JArray);
                    g.Workflow.Remove(id);
                }
            });
            g.RunOnNodesOfClass("LTXVSeparateAVLatent", (id, data) =>
            {
                JArray source = data["inputs"]["av_latent"] as JArray;
                string sourceNode = $"{source[0]}";
                JObject actualNode = g.Workflow[sourceNode] as JObject;
                if ($"{actualNode["class_type"]}" == "LTXVConcatAVLatent")
                {
                    g.ReplaceNodeConnection([id, 0], actualNode["inputs"]["video_latent"] as JArray);
                    g.ReplaceNodeConnection([id, 1], actualNode["inputs"]["audio_latent"] as JArray);
                }
            });
            void fixDecode(string id, JObject data)
            {
                JArray source = data["inputs"]["samples"] as JArray;
                string sourceNode = $"{source[0]}";
                JObject actualNode = g.Workflow[sourceNode] as JObject;
                JArray myVae = data["inputs"]["vae"] as JArray;
                if ($"{actualNode["class_type"]}" == "LTXVSeparateAVLatent")
                {
                    JArray subSource = actualNode["inputs"]["av_latent"] as JArray;
                    string subSourceNode = $"{subSource[0]}";
                    JObject actualSubNode = g.Workflow[subSourceNode] as JObject;
                    if ($"{actualSubNode["class_type"]}" == "VAEEncode" || $"{actualSubNode["class_type"]}" == "VAEEncodeTiled")
                    {
                        JArray srcVae = actualSubNode["inputs"]["vae"] as JArray;
                        if ($"{myVae[0]}" == $"{srcVae[0]}" && $"{myVae[1]}" == $"{srcVae[1]}")
                        {
                            JArray srcImage = actualSubNode["inputs"]["pixels"] as JArray;
                            g.ReplaceNodeConnection([id, 0], srcImage);
                            g.Workflow.Remove(id);
                        }
                    }
                }
                else if ($"{actualNode["class_type"]}" == "VAEEncode" || $"{actualNode["class_type"]}" == "VAEEncodeTiled")
                {
                    JArray srcVae = actualNode["inputs"]["vae"] as JArray;
                    if ($"{myVae[0]}" == $"{srcVae[0]}" && $"{myVae[1]}" == $"{srcVae[1]}")
                    {
                        JArray srcImage = actualNode["inputs"]["pixels"] as JArray;
                        g.ReplaceNodeConnection([id, 0], srcImage);
                        g.Workflow.Remove(id);
                    }
                }
            }
            void fixEncode(string id, JObject data)
            {
                JArray source = data["inputs"]["pixels"] as JArray;
                string sourceNode = $"{source[0]}";
                JObject actualNode = g.Workflow[sourceNode] as JObject;
                JArray myVae = data["inputs"]["vae"] as JArray;
                // TODO: LTX concat
                if ($"{actualNode["class_type"]}" == "VAEDecode" || $"{actualNode["class_type"]}" == "VAEDecodeTiled")
                {
                    JArray srcVae = actualNode["inputs"]["vae"] as JArray;
                    if ($"{myVae[0]}" == $"{srcVae[0]}" && $"{myVae[1]}" == $"{srcVae[1]}")
                    {
                        JArray srcLatent = actualNode["inputs"]["samples"] as JArray;
                        g.ReplaceNodeConnection([id, 0], srcLatent);
                        g.Workflow.Remove(id);
                    }
                }
            }
            g.RunOnNodesOfClass("VAEDecode", fixDecode);
            g.RunOnNodesOfClass("VAEDecodeTiled", fixDecode);
            g.RunOnNodesOfClass("VAEEncode", fixEncode);
            g.RunOnNodesOfClass("VAEEncodeTiled", fixEncode);
            g.RemoveClassesIfUnused(AutoCleanupNodeTypes);
        }, 200);
        #endregion
    }

    public static HashSet<string> AutoCleanupNodeTypes =
    [
        "VAEDecode", "VAEDecodeTiled", "VAEEncode", "CLIPTextEncode", "CLIPTextEncodeSDXL",
        "LTXVAudioVAEDecode", "LTXVSeparateAVLatent", "LTXVConditioning", "LTXVEmptyLatentAudio", "LTXVConcatAVLatent", "LTXVReferenceAudio",
        "SwarmCountFrames", "SwarmClipTextEncodeAdvanced"
    ];
}