huggingface / huggingface/diffusers
[Feature] Support Unipic 3.0 Pipeline
- Dominant language
- Python
- Stars
- 34.5k
- Forks
- 7.3k
- Avg merge
- 3d 3h
- Merged PRs (30d)
- 91
Description
**Is your feature request related to a problem? Please describe.**
Yes. Currently, there is no native pipeline in diffusers that supports Unified Multi-Image Conditioning as proposed in UniPic 3.0. https://huggingface.co/Skywork/Unipic3
**Describe the solution you'd like.**
I propose adding a UniPic 3.0–style pipeline to diffusers that natively supports multi-image, multi-type conditioning within a single diffusion forward pass.
**Describe alternatives you've considered.**
Extending existing pipelines: Results in model-specific hacks rather than a reusable abstraction. And it's difficult to maintain or generalize.
**Additional context.**
Add any other context or screenshots about the feature request here.
Contributor guide
Research direction
Start by reviewing the UniPic 3.0 model page linked in the issue and the existing diffusers pipeline patterns. Define how multi-image, multi-type conditioning should work within one diffusion forward pass. Done means diffusers provides a native, reusable UniPic 3.0-style pipeline rather than model-specific extensions.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100