MuseV
TMElyralab/MuseVReviewed by AI
Checked by RepoPicks Radar on
What is MuseV?
MuseV, from Tencent Music's Lyra Lab, makes videos of virtual people from a photo, a text prompt or another video, and can keep going for long clips. It is a research release: you run it by cloning the repository and typing commands, ideally inside the project's Docker image, on a machine with an NVIDIA card (the README mentions about 8 to 12 GB of video memory). Model files are downloaded separately from Hugging Face. The code is under the MIT license, but the trained models are stated to be for non-commercial research only.
MuseV is free and open source (Apache-2.0) and can replace Runway, HeyGen, and D-ID.
How easy is it to install MuseV?
D · terminal, several tools or a specific graphics card · Windows
- Windows D
- Linux D
On Linux, the recommended path is pulling the Docker image, cloning the repository with submodules, setting PYTHONPATH by hand, downloading model files and running Python commands with many flags, so it is a D. Windows is the same route through Docker with an NVIDIA card; the README also links a third-party one-click Windows package but says third-party integrations are not verified or maintained by the authors.
Coming soon: install in 1 click with the RepoPicks app.
MuseV on GitHub
- Stars
- 2,851on Nov 10, 2026
- Downloads
- —all releases
- Growth
- +1stars in 6 days
Stars over time
Facts
- Runs on
- Linux, Windows
- You need to install
- Python, Docker, NVIDIA graphics card, FFmpeg
- To build from source
- Python, Docker, NVIDIA graphics card, FFmpeg
- Disk space
- 1.6 GB–6.2 GB · estimate
- Accounts or keys
- None
- Cost
- Free · No payment to use the code; the README states the trained models are for non-commercial research purposes only.
- License
- Apache-2.0 (Permissive)
- Last change on GitHub
- May 29, 2024
MuseV risk signals (8)
- AttentionDependencies with known vulnerabilities
Pinned dependencies with known vulnerabilities (OSV): transformers: GHSA-29pf-2h5f-8g72, GHSA-37mw-44qp-f5jm, GHSA-37q5-v5qm-c9v8; accelerate: GHSA-4j2p-28q2-5m79, PYSEC-2026-3804; onnx: GHSA-3r9x-f23j-gc73, GHSA-48m4-24cq-37qf, GHSA-538c-55jv-c5g9; protobuf: GHSA-7gcm-g887-7qv7, GHSA-8qvm-5x2c-j2w7, PYSEC-2026-1805; opencv-contrib-python: GHSA-cxjf-x6jp-p7mc, PYSEC-2023-181; pytorch-lightning: GHSA-4cv3-v7pv-rfhf, GHSA-75m9-98v2-hjpm, GHSA-98fp-7v67-4v3q; gdown: GHSA-76hw-p97h-883f, PYSEC-2026-2158; redis: GHSA-24wv-mv5m-xv4h, GHSA-8fww-64cx-x8p5, PYSEC-2023-45
automatic check
- Attentionmodel license non commercial
README.md, Disclaimer/License: "model: The trained model are available for non-commercial research purposes only." and "The testdata are collected from internet, which are available for non-commercial research purposes only."
AI review
- Attentionthird party download link
README.md points to an unofficial Windows package on a file-sharing site ("One click integration package in windows", netdisk link with code "glut") and says: "we have not verified, maintained, or updated third-party".
AI review
- Infosetup runs installs
setup.py runs package installs while the project is being installed: subprocess.run(["pip", "install", "basicsr"]) plus mim install of mmengine, mmcv, mmdet and mmpose.
AI review
- Infodependencies from git branches
requirements.txt pulls code straight from moving branches: "diffusers @ git+https://github.com/TMElyralab/diffusers.git@tme", "ip_adapter @ git+https://github.com/tencent-ailab/IP-Adapter.git@main", "clip @ git+https://github.com/openai/CLIP.git@main", so what gets installed can change over time.
AI review
- Infolicense file mismatch
The LICENSE file starts with "MIT License Copyright (c) 2024 Tencent Music Entertainment Group" and then copies the Apache 2.0 text for bundled dependencies; the README says "The code of MuseV is released under the MIT License".
AI review
- Infogenerated people videos
README.md shows generated videos of real and famous faces (for example Mona Lisa and a named influencer, "Sun Xinying is a supermodel KOL") and the disclaimer says users "are expected to comply with local laws" and that the developers take no responsibility for misuse.
AI review
- Infoproject quality warnings by authors
README.md Limitation section: "Lack of generalization ability", "Watermarks may appear because of webvid", "Understructured code... complex and unrefacted codes".
AI review
Evidence is shown in its original language.
What Radar could not check about MuseV
- Whether it still installs today: the last commit was in May 2024 and the pinned versions (torch 2.0.1, transformers 4.33.1) are old.
- Real disk space and video memory needed after downloading the model files from Hugging Face.
- Whether the Hugging Face demo space linked in the README is still online.
- Whether the third-party Windows one-click package and ComfyUI node work or match this code.
- Video quality for your own photos: the authors say results vary a lot with the image and the base model used.