HiDream-O1 is a Qwen3-VL based image generation model that predicts raw RGB image patches directly. Unlike HiDream-I1, it does not use a VAE component.
The following model is available for the [HiDreamO1ImagePipeline] pipeline:
| Model | Hugging Face Hub |
|---|---|
| HiDream-O1-Image | HiDream-ai/HiDream-O1-Image |
| HiDream-O1-Image-Dev | HiDream-ai/HiDream-O1-Image-Dev |
[[autodoc]] HiDreamO1ImagePipeline