Open-Sora

Apache-2.0 research stack for training and running video diffusion on 24 GB cards.

Install Repository Compare

Fit.

Catalog snapshot

Platforms: Linux, Macos, Windows
Operating systems: Linux, macOS, Windows
Ram floor: 32 GB
Gpu: Required · 24 GB VRAM
Docker: No
Setup: Advanced
Deploy as: Local
Commercial use: Allowed

Built for teams fine-tuning video models, not end users rendering clips.

Project.

Fetched Sep 12, 2026¹

License: Apache 2.0
Stars: 29.8k¹
Forks: 3.1k¹
Open issues: 0¹
Language: Python¹
Last commit: Apr 9, 2026¹
Tags: videotrainingapache

About.

Open-Sora is the research stack for video generation: model, data pipeline, and training recipes in one Apache-2.0 repo, with a 24 GB consumer-GPU path. Quality trails Hunyuan and Wan. The value is the training code, not the outputs.

full training pipeline · data preprocessing · 24 GB VRAM path · Apache-2.0

Alternatives.

Nearby.

All Video AI

WWan2.1Alibaba’s open text-to-video and image-to-video stack.Compare
CCogVideoXTHUDM open video diffusion with strong prompt following.Compare
FFramePackLong local image-to-video with a low VRAM floor.Compare
LLTX-VideoLightricks open video model aimed at faster local clips.Compare
MMochiGenmo’s Apache-2.0 10B video model — permissive weights, heavy hardware.Compare