Open-Sora

What it does

Commercial use

Yeschecked 19 Sep 2026

Compute requirements

The README documents 768x768 generation using multi-GPU inference and reports benchmarks on H100/H800 GPUs.checked 13 Oct 2026

Generation modes

The 11B model supports text-to-video and image-to-video generation.checked 12 Oct 2026

Generation settings

Users can adjust aspect ratio and frame count; the documented frame counts follow 4k+1 and must be less than 129.checked 13 Oct 2026

Hardware

The documented efficiency measurements use H100 and H800 GPUs, and 768x768 inference examples include multi-GPU execution.checked 12 Oct 2026

Image-to-video

Yeschecked 19 Sep 2026

Image-to-video input

Image-to-video generation accepts a prompt and a reference image.checked 12 Oct 2026

Installation

The quickstart uses Python 3.10 and requires PyTorch 2.4.0 or later.checked 13 Oct 2026

License

The repository identifies its license as Apache-2.0.checked 13 Oct 2026

Model downloads

The README provides model download instructions for Hugging Face and ModelScope.checked 13 Oct 2026

Model size

The repository describes Open-Sora 2.0 as an 11B model.checked 12 Oct 2026

Motion scoring

Inference supports a motion score, and a dynamic motion-score evaluator is available after setting an OpenAI API key.checked 12 Oct 2026

Project audience

The project says it aims to make video-generation models, tools, and implementation details accessible to all.checked 13 Oct 2026

Prompt refinement

The project can use ChatGPT to refine prompts for text-to-video and image-to-video generation, with an OpenAI API key configured.checked 12 Oct 2026

Prompt tools

The project documents prompt refinement with ChatGPT and dynamic motion-score evaluation using an OpenAI API key.checked 13 Oct 2026

Purpose

Open-Sora is an open-source project for producing high-quality video.checked 13 Oct 2026

Requirements

The quickstart uses Python 3.10 and requires PyTorch 2.4.0 or later.checked 12 Oct 2026

Resolutions

The 11B model supports 256px and 768px resolutions.checked 13 Oct 2026

Support

The repository provides a contribution guideline for people wishing to contribute.checked 13 Oct 2026

Text-to-video

Yeschecked 19 Sep 2026

Text-to-video pipeline

For high-quality text-to-video, the project provides a text-to-image-to-video pipeline using the Flux text-to-image model.checked 12 Oct 2026

Training resources

The repository links reports and instructions for training or fine-tuning a model and training and evaluating a video autoencoder.checked 13 Oct 2026

Versions

The README says Open-Sora is under active development and maintains separate branches for different versions.checked 12 Oct 2026

Video generation

The 11B model supports text-to-video and image-to-video generation.checked 13 Oct 2026