# kohya-ss/musubi-tuner

Repository: https://github.com/kohya-ss/musubi-tuner
Canonical: https://ross.abutalabs.com/products/musubi-tuner
Language: Python
License Family: other
Last push: 2026-08-25T11:11:10+00:00

## Health v2 (maintenance only)
Score: 85/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 90, longevity 43
- inputs: {"age_days": 610, "days_push": 8, "days_rel": 71, "gap_med": 3.5, "n_releases_24m": 3}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2002, forks 306 (observed 2026-08-28T04:06:04.422580+00:00)

## What it is
Musubi Tuner is a set of Python scripts for training LoRA (Low-Rank Adaptation) adapters for video and image generation model architectures such as HunyuanVideo, Wan2.1/2.2, FramePack, FLUX.1 Kontext, FLUX.2, Qwen-Image, and Z-Image. It handles dataset configuration, latent/text-encoder caching, accelerated training, and inference with the fine-tuned models.

## Use cases
- train a lora for hunyuanvideo video generation
- fine-tune wan 2.1 with my own video clips
- train a lora on qwen-image
- low-rank adaptation training for flux.1 kontext
- cache latents and train a video model on a single gpu
- fine-tune image generation models with custom datasets

## When to choose
- you want to train LoRA adapters for supported video/image diffusion architectures
- you need GPU-efficient fine-tuning with pre-caching of latents and text embeddings
- you are comfortable with command-line training scripts and accelerate configuration

## When to avoid
- you need a GUI-based training tool
- you want to fine-tune architectures not on the supported list
- you need an officially supported training pipeline from the model authors

## Facets
- artifact type: library
- maturity: active
- function: llm-training, machine-learning, deep-learning, cli
- domain: machine-learning, deep-learning, image-processing, artificial-intelligence
- platform: windows, python
- tags: lora-training, fine-tuning, hunyuanvideo, wan, flux, qwen-image, diffusion-models, video-generation, video, linux, macos, gpu

## Member repositories
- kohya-ss/musubi-tuner (main) score 85

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:04.422580+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 2026-08-30T03:01:56.950636+00:00, confidence not recorded.
  - readme: https://github.com/kohya-ss/musubi-tuner (fetched 2026-08-28T04:06:04.422580+00:00, sha 706504786ee3)
- Data as of 2026-08-30T08:39:29.467469+00:00.