# ChenyangSi/FreeU

FreeU: Free Lunch in Diffusion U-Net (CVPR2024 Oral)

Repository: https://github.com/ChenyangSi/FreeU
Canonical: https://ross.abutalabs.com/products/freeu
Homepage: https://chenyangsi.top/FreeU/
License: MIT
License Family: permissive
Last push: 2024-12-24T02:10:30+00:00

## Health v2 (maintenance only)
Score: 28/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 77
- inputs: {"age_days": 1084, "days_push": 618, "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 1899, forks 76 (observed 2026-08-28T04:05:51.394879+00:00)

## What it is
FreeU is a research codebase implementing a training-free method that improves diffusion model sample quality by re-weighting U-Net backbone and skip-connection features via Fourier filtering. It integrates with models like Stable Diffusion, DreamBooth, and ModelScope with only two scaling factors adjusted at inference.

## Use cases
- improve stable diffusion image quality without retraining
- speed up diffusion sampling quality for free
- apply FreeU scaling factors to SDXL or ControlNet
- enhance text-to-image and video generation output
- reproduce CVPR 2024 FreeU paper results

## When to choose
- you use diffusion models like Stable Diffusion and want better samples with no training or latency cost
- you want a few-lines-of-code drop-in enhancement for existing diffusion pipelines

## When to avoid
- you need a production-maintained library with ongoing support
- your pipeline does not use a U-Net-based diffusion architecture

## Facets
- artifact type: library
- maturity: maintenance
- function: image-processing, machine-learning, deep-learning
- domain: image-processing, deep-learning, artificial-intelligence
- platform: python, cross-platform
- tags: diffusion-models, stable-diffusion, u-net, text-to-image, video-generation, research-code, cvpr-2024, gpu

## Member repositories
- ChenyangSi/FreeU (main) score 28

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:51.394879+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 2026-08-30T03:12:05.538951+00:00, confidence not recorded.
  - readme: https://github.com/ChenyangSi/FreeU (fetched 2026-08-28T04:05:51.394879+00:00, sha 5450eb653780)
  - homepage: https://chenyangsi.top/FreeU/ (fetched 2026-08-29T10:51:40.906815+00:00, sha 593bfc61c2db)
- Data as of 2026-08-30T08:39:29.467469+00:00.