Vidéo
HunyuanVideo
HunyuanVideo: A Systematic Framework For Large Video Generation Model
★ 12,6 k étoilesCustom (check repo)
Voir le guide →Mochi 1 est un projet de vidéo open source, créé par genmoai (genmoai/mochi sur GitHub). Il compte 3 755 étoiles sur GitHub et 494 forks, il est publié sous la licence Apache-2.0 et a été mis à jour pour la dernière fois sur GitHub le 22 octobre 2026.
Mochi 1 — Genmo's open state-of-the-art video generation model, with high-fidelity motion and strong prompt adherence, Apache-2.0
Clone, create a Python 3.11 virtual environment with uv, activate it, and sync the locked dependencies.
Download the model and VAE into a local weights directory with the bundled script (or from Hugging Face).
Run generation, and make sure ffmpeg is installed — it turns the model outputs into viewable videos.
High-fidelity motion and strong prompt adherence, released openly instead of behind an API.
Code and weights under Apache-2.0, so the model can be used and built on freely.
Fine-tune Mochi 1 on your own style or subject with the bundled fine-tuner.
Native ComfyUI support lets smaller GPUs run the model where the full pipeline cannot.
The README install route with uv (Python 3.11+).
git clone https://github.com/genmoai/mochi
cd mochi
pip install uv
uv venv --python 3.11 .venv
source .venv/bin/activate
uv sync --lockedThe bundled download script for the model and VAE.
python3 ./scripts/download_weights.py weights/The hosted playground on Genmo's site, before installing anything.
https://www.genmo.ai/playGenmo, as its open state-of-the-art video generation model.
Yes — Apache-2.0 licensed, including the weights.
The full pipeline wants a strong GPU, but consumer-GPU support exists natively in ComfyUI, and the playground lets you try it free first.
Python 3.11+, uv, the model weights, and ffmpeg for turning outputs into videos.
Clone, uv sync the locked environment, download the weights, and generate — or try the Genmo playground first.
HunyuanVideo: A Systematic Framework For Large Video Generation Model
Real-time long video generation — a research model built for speed.
Research model enforcing 3D constraints in text-to-video generation.
LiveTalk is a unified, high-performance talking head generation system that combines
The benchmark suite for evaluating AI video generation models.
Time-lapse video generation that simulates how the world changes.
Les autres projets de la catégorie Vidéo avec le plus d'étoiles du catalogue LibreAI — la sélection réaliste d'alternatives si Mochi 1 ne vous convient pas.
A feature-rich command-line audio/video downloader
real time face swap and one-click video deepfake with only a single image
OBS Studio - Free and open source software for live streaming and screen recording
Mirror of https://git.ffmpeg.org/ffmpeg.git
World's first open-source, agentic video production system. 12 production pipelines, 100+ tools, 700+ agent skill and production-knowledge files. Turn your A…
Projets de la catégorie Vidéo proches dans le catalogue, les plus proches en étoiles de Mochi 1 — continuez à explorer de part et d'autre.
PRINTFILM turns scripts and topics into finished AI short videos and animated short dramas
Making ComfyUI more comfortable!
Open-source video compositing software. Node-graph based. Similar in functionalities to Adobe After Effects and Nuke by The Foundry.
An AI video editing agent — describe the story, and it plans and cuts the video for you.
OpenShot Video Editor is an award-winning free and open-source video editor for Linux, Mac, and Windows, and is dedicated to delivering high quality video ed…
An MCP server that lets AI agents drive DaVinci Resolve Studio — editing, grading and more.
Gratuit et ouvert : récupérez le code source, mettez une étoile au dépôt et commencez à créer.