# Phantom-video/HuMo

HuMo: Human-Centric Video Generation via Collaborative Multi-Modal Conditioning

Repository: https://github.com/Phantom-video/HuMo
Canonical: https://ross.abutalabs.com/products/humo
Homepage: https://phantom-video.github.io/HuMo/
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-01-25T17:14:34+00:00

## Health v2 (maintenance only)
Score: 46/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 64, release rhythm 35, longevity 25
- inputs: {"age_days": 358, "days_push": 220, "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 1283, forks 246 (observed 2026-08-28T04:04:14.426199+00:00)

## What it is
HuMo is a research model and Python codebase from Tsinghua University and ByteDance for human-centric video generation using collaborative multi-modal conditioning (text, reference image, and audio). It ships 1.7B and 17B model weights that generate subject-consistent, audio-visual-synchronized talking videos.

## Use cases
- generate talking-head videos from a photo and audio
- create lip-synced avatar videos from text prompts
- generate videos that keep a reference person's identity consistent
- run video generation locally on a single GPU
- train or fine-tune human-centric video generation models
- use video generation inside ComfyUI workflows

## When to choose
- you need subject-consistent human video generation with audio-visual sync
- you want open weights runnable on a single 32G GPU (1.7B model)
- you need a research-grade multimodal video generation model with dataset access

## When to avoid
- you need fast real-time video generation
- you want a polished end-user app rather than a research codebase
- you lack a GPU - the 17B model has heavy hardware requirements

## Facets
- artifact type: library
- maturity: active
- function: video-processing, machine-learning, deep-learning, speech-recognition, llm-inference
- domain: artificial-intelligence, deep-learning, computer-vision
- platform: python
- tags: video-generation, talking-head, audio-visual-sync, diffusion-model, subject-consistency, multimodal, comfyui, huggingface, video, audio, gpu, linux, docker

## Member repositories
- Phantom-video/HuMo (main) score 46

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:14.426199+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 2026-08-30T04:56:47.396582+00:00, confidence not recorded.
  - readme: https://github.com/Phantom-video/HuMo (fetched 2026-08-28T04:04:14.426199+00:00, sha 20ac0aa35510)
  - homepage: https://phantom-video.github.io/HuMo/ (fetched 2026-08-29T12:12:39.961625+00:00, sha e4dc15ee2a53)
- Data as of 2026-08-30T08:39:29.467469+00:00.