3131
3232# auto_docstring
3333class HunyuanVideo15CoreDenoiseStep (SequentialPipelineBlocks ):
34+ """
35+ Denoise block that takes encoded conditions and runs the denoising process.
36+
37+ Components:
38+ transformer (`HunyuanVideo15Transformer3DModel`)
39+ scheduler (`FlowMatchEulerDiscreteScheduler`)
40+ guider (`ClassifierFreeGuidance`)
41+
42+ Inputs:
43+ num_videos_per_prompt (`None`, *optional*, defaults to 1):
44+ TODO: Add description.
45+ prompt_embeds (`Tensor`):
46+ TODO: Add description.
47+ batch_size (`int`, *optional*):
48+ TODO: Add description.
49+ num_inference_steps (`None`, *optional*, defaults to 50):
50+ TODO: Add description.
51+ sigmas (`None`, *optional*):
52+ TODO: Add description.
53+ height (`int`, *optional*):
54+ TODO: Add description.
55+ width (`int`, *optional*):
56+ TODO: Add description.
57+ num_frames (`int`, *optional*, defaults to 121):
58+ TODO: Add description.
59+ latents (`Tensor | NoneType`, *optional*):
60+ TODO: Add description.
61+ generator (`None`, *optional*):
62+ TODO: Add description.
63+ attention_kwargs (`None`, *optional*):
64+ TODO: Add description.
65+ negative_prompt_embeds (`Tensor`, *optional*):
66+ TODO: Add description.
67+ prompt_embeds_mask (`Tensor`):
68+ TODO: Add description.
69+ negative_prompt_embeds_mask (`Tensor`, *optional*):
70+ TODO: Add description.
71+ prompt_embeds_2 (`Tensor`):
72+ TODO: Add description.
73+ negative_prompt_embeds_2 (`Tensor`, *optional*):
74+ TODO: Add description.
75+ prompt_embeds_mask_2 (`Tensor`):
76+ TODO: Add description.
77+ negative_prompt_embeds_mask_2 (`Tensor`, *optional*):
78+ TODO: Add description.
79+
80+ Outputs:
81+ latents (`Tensor`):
82+ Denoised latents.
83+ """
84+
3485 model_name = "hunyuan-video-1.5"
3586 block_classes = [
3687 HunyuanVideo15TextInputStep ,
@@ -51,6 +102,69 @@ def outputs(self):
51102
52103# auto_docstring
53104class HunyuanVideo15Blocks (SequentialPipelineBlocks ):
105+ """
106+ Modular pipeline blocks for HunyuanVideo 1.5 text-to-video.
107+
108+ Components:
109+ text_encoder (`Qwen2_5_VLTextModel`)
110+ tokenizer (`Qwen2TokenizerFast`)
111+ text_encoder_2 (`T5EncoderModel`)
112+ tokenizer_2 (`ByT5Tokenizer`)
113+ guider (`ClassifierFreeGuidance`)
114+ transformer (`HunyuanVideo15Transformer3DModel`)
115+ scheduler (`FlowMatchEulerDiscreteScheduler`)
116+ vae (`AutoencoderKLHunyuanVideo15`)
117+ video_processor (`HunyuanVideo15ImageProcessor`)
118+
119+ Inputs:
120+ prompt (`None`, *optional*):
121+ TODO: Add description.
122+ negative_prompt (`None`, *optional*):
123+ TODO: Add description.
124+ prompt_embeds (`Tensor`, *optional*):
125+ TODO: Add description.
126+ prompt_embeds_mask (`Tensor`, *optional*):
127+ TODO: Add description.
128+ negative_prompt_embeds (`Tensor`, *optional*):
129+ TODO: Add description.
130+ negative_prompt_embeds_mask (`Tensor`, *optional*):
131+ TODO: Add description.
132+ prompt_embeds_2 (`Tensor`, *optional*):
133+ TODO: Add description.
134+ prompt_embeds_mask_2 (`Tensor`, *optional*):
135+ TODO: Add description.
136+ negative_prompt_embeds_2 (`Tensor`, *optional*):
137+ TODO: Add description.
138+ negative_prompt_embeds_mask_2 (`Tensor`, *optional*):
139+ TODO: Add description.
140+ num_videos_per_prompt (`int`, *optional*, defaults to 1):
141+ TODO: Add description.
142+ batch_size (`int`, *optional*):
143+ TODO: Add description.
144+ num_inference_steps (`None`, *optional*, defaults to 50):
145+ TODO: Add description.
146+ sigmas (`None`, *optional*):
147+ TODO: Add description.
148+ height (`int`, *optional*):
149+ TODO: Add description.
150+ width (`int`, *optional*):
151+ TODO: Add description.
152+ num_frames (`int`, *optional*, defaults to 121):
153+ TODO: Add description.
154+ latents (`Tensor | NoneType`, *optional*):
155+ TODO: Add description.
156+ generator (`None`, *optional*):
157+ TODO: Add description.
158+ attention_kwargs (`None`, *optional*):
159+ TODO: Add description.
160+ output_type (`str`, *optional*, defaults to np):
161+ TODO: Add description.
162+
163+ Outputs:
164+ videos (`list`):
165+ The generated videos.
166+ """
167+
54168 model_name = "hunyuan-video-1.5"
55169 block_classes = [
56170 HunyuanVideo15TextEncoderStep ,
@@ -70,6 +184,59 @@ def outputs(self):
70184
71185# auto_docstring
72186class HunyuanVideo15Image2VideoCoreDenoiseStep (SequentialPipelineBlocks ):
187+ """
188+ Denoise block for image-to-video that takes encoded conditions and runs the denoising process.
189+
190+ Components:
191+ transformer (`HunyuanVideo15Transformer3DModel`)
192+ scheduler (`FlowMatchEulerDiscreteScheduler`)
193+ vae (`AutoencoderKLHunyuanVideo15`)
194+ video_processor (`HunyuanVideo15ImageProcessor`)
195+ image_encoder (`SiglipVisionModel`)
196+ feature_extractor (`SiglipImageProcessor`)
197+ guider (`ClassifierFreeGuidance`)
198+
199+ Inputs:
200+ num_videos_per_prompt (`None`, *optional*, defaults to 1):
201+ TODO: Add description.
202+ prompt_embeds (`Tensor`):
203+ TODO: Add description.
204+ batch_size (`int`, *optional*):
205+ TODO: Add description.
206+ num_inference_steps (`None`, *optional*, defaults to 50):
207+ TODO: Add description.
208+ sigmas (`None`, *optional*):
209+ TODO: Add description.
210+ image (`None`):
211+ TODO: Add description.
212+ num_frames (`int`, *optional*, defaults to 121):
213+ TODO: Add description.
214+ latents (`Tensor | NoneType`, *optional*):
215+ TODO: Add description.
216+ generator (`None`, *optional*):
217+ TODO: Add description.
218+ attention_kwargs (`None`, *optional*):
219+ TODO: Add description.
220+ negative_prompt_embeds (`Tensor`, *optional*):
221+ TODO: Add description.
222+ prompt_embeds_mask (`Tensor`):
223+ TODO: Add description.
224+ negative_prompt_embeds_mask (`Tensor`, *optional*):
225+ TODO: Add description.
226+ prompt_embeds_2 (`Tensor`):
227+ TODO: Add description.
228+ negative_prompt_embeds_2 (`Tensor`, *optional*):
229+ TODO: Add description.
230+ prompt_embeds_mask_2 (`Tensor`):
231+ TODO: Add description.
232+ negative_prompt_embeds_mask_2 (`Tensor`, *optional*):
233+ TODO: Add description.
234+
235+ Outputs:
236+ latents (`Tensor`):
237+ Denoised latents.
238+ """
239+
73240 model_name = "hunyuan-video-1.5"
74241 block_classes = [
75242 HunyuanVideo15TextInputStep ,
@@ -90,6 +257,69 @@ def outputs(self):
90257
91258# auto_docstring
92259class HunyuanVideo15Image2VideoBlocks (SequentialPipelineBlocks ):
260+ """
261+ Modular pipeline blocks for HunyuanVideo 1.5 image-to-video.
262+
263+ Components:
264+ text_encoder (`Qwen2_5_VLTextModel`)
265+ tokenizer (`Qwen2TokenizerFast`)
266+ text_encoder_2 (`T5EncoderModel`)
267+ tokenizer_2 (`ByT5Tokenizer`)
268+ guider (`ClassifierFreeGuidance`)
269+ transformer (`HunyuanVideo15Transformer3DModel`)
270+ scheduler (`FlowMatchEulerDiscreteScheduler`)
271+ vae (`AutoencoderKLHunyuanVideo15`)
272+ video_processor (`HunyuanVideo15ImageProcessor`)
273+ image_encoder (`SiglipVisionModel`)
274+ feature_extractor (`SiglipImageProcessor`)
275+
276+ Inputs:
277+ prompt (`None`, *optional*):
278+ TODO: Add description.
279+ negative_prompt (`None`, *optional*):
280+ TODO: Add description.
281+ prompt_embeds (`Tensor`, *optional*):
282+ TODO: Add description.
283+ prompt_embeds_mask (`Tensor`, *optional*):
284+ TODO: Add description.
285+ negative_prompt_embeds (`Tensor`, *optional*):
286+ TODO: Add description.
287+ negative_prompt_embeds_mask (`Tensor`, *optional*):
288+ TODO: Add description.
289+ prompt_embeds_2 (`Tensor`, *optional*):
290+ TODO: Add description.
291+ prompt_embeds_mask_2 (`Tensor`, *optional*):
292+ TODO: Add description.
293+ negative_prompt_embeds_2 (`Tensor`, *optional*):
294+ TODO: Add description.
295+ negative_prompt_embeds_mask_2 (`Tensor`, *optional*):
296+ TODO: Add description.
297+ num_videos_per_prompt (`int`, *optional*, defaults to 1):
298+ TODO: Add description.
299+ batch_size (`int`, *optional*):
300+ TODO: Add description.
301+ num_inference_steps (`None`, *optional*, defaults to 50):
302+ TODO: Add description.
303+ sigmas (`None`, *optional*):
304+ TODO: Add description.
305+ image (`None`):
306+ TODO: Add description.
307+ num_frames (`int`, *optional*, defaults to 121):
308+ TODO: Add description.
309+ latents (`Tensor | NoneType`, *optional*):
310+ TODO: Add description.
311+ generator (`None`, *optional*):
312+ TODO: Add description.
313+ attention_kwargs (`None`, *optional*):
314+ TODO: Add description.
315+ output_type (`str`, *optional*, defaults to np):
316+ TODO: Add description.
317+
318+ Outputs:
319+ videos (`list`):
320+ The generated videos.
321+ """
322+
93323 model_name = "hunyuan-video-1.5"
94324 block_classes = [
95325 HunyuanVideo15TextEncoderStep ,
0 commit comments