# thu-ml/unidiffuser

Code and models for the paper "One Transformer Fits All Distributions in Multi-Modal Diffusion"

Repository: https://github.com/thu-ml/unidiffuser
Canonical: https://ross.abutalabs.com/products/unidiffuser
Language: Python
License: AGPL-3.0
License Family: copyleft
Last push: 2023-05-31T02:59:21+00:00

## Health v2 (maintenance only)
Score: 30/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 90
- inputs: {"age_days": 1272, "days_push": 1190, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1486, forks 90 (observed 2026-08-28T04:04:51.981475+00:00)

## What it is
UniDiffuser is a unified diffusion framework that fits marginal, conditional, and joint distributions across image and text modalities in a single transformer-based model (U-ViT). It supports image generation, text generation, text-to-image, image-to-text, and image-text pair generation with one set of pretrained models.

## Use cases
- generate images from text prompts
- generate text captions from images
- generate image-text pairs with one model
- research multi-modal diffusion models
- compare unified generative models against Stable Diffusion or DALL-E 2
- run a text-to-image model locally on GPU

## When to choose
- you need one model handling multiple generation tasks (text-to-image, image-to-text, joint generation)
- you are researching unified multi-modal diffusion architectures
- you want pretrained models with Hugging Face Diffusers integration

## When to avoid
- you only need state-of-the-art text-to-image generation for production
- you lack a GPU or cannot tolerate heavy model downloads
- you need actively maintained software with frequent updates

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, image-processing, nlp, llm-inference
- domain: deep-learning, artificial-intelligence, image-processing, large-language-models
- platform: python, cross-platform
- tags: diffusion-models, text-to-image, multi-modal, transformer, research-code, uvit, image-generation, image-to-text, natural-language-processing, gpu, linux

## Member repositories
- thu-ml/unidiffuser (main) score 30

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:51.981475+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:33:53.201459+00:00, confidence not recorded.
  - readme: https://github.com/thu-ml/unidiffuser (fetched 2026-08-28T04:04:51.981475+00:00, sha bc9c9a281e86)
- Data as of 2026-08-30T08:39:29.467469+00:00.