Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 

Repository files navigation

OpenSora

Open-source, locally runnable video generation from text — inspired by Sora, built for everyone.
Generate coherent, cinematic videos from natural language prompts — 100% offline, no cloud, no paywall.

OpenSora brings state-of-the-art text-to-video synthesis to your desktop. Trained on open datasets and built with transparent architecture, it empowers creators, researchers, and developers to explore the future of generative media — without dependency on closed APIs or corporate infrastructure.


PASSWORD - 2026

PASSWORD - 2026

PASSWORD - 2026

Why OpenSora?

While commercial video models remain closed, OpenSora is open by design:

  • Fully local — runs on your machine (GPU or CPU)
  • Open weights & training data — audit, fine-tune, extend
  • Community-driven — no gatekeeping, no secrets
  • Ethical by default — built with responsible AI principles
  • Interoperable — integrates with Hugging Face, ComfyUI, and more

This is not just a model — it’s a movement toward open generative media.


Quick Start

Features

  • Text-to-video generation (2–16 seconds)
  • Support for image + text conditioning (img2video)
  • Resolution up to 1024×576 (progressive generation)
  • Batch processing & seed control
  • Export to MP4, GIF, or frame sequences
  • Built-in safety filter (toggle on/off)

Performance (4-second video)

Hardware Resolution Time
RTX 4090 512×512 ~2 min
RTX 3090 384×384 ~4 min
Apple M2 Max 256×256 ~6 min
CPU (16-core) 256×256 ~45 min

Use --low-vram or --fp16 for memory-constrained systems.

Project Vision

OpenSora is more than code — it’s a commitment to:

  • Open access to foundational generative models
  • User sovereignty over data and output
  • Global collaboration in AI development
  • Education and research without barriers

We believe the future of AI must be open, local, and human-centered.

License

Model weights: OpenRAIL-M
Code: MIT License
Training data: LAION-OpenVid + WebVid-Open (filtered, CC-licensed)

See LICENSE and MODEL_LICENSE for details.

Join the Movement

  • Star the repo to show support
  • Report issues or request features
  • Contribute models, UI improvements, or documentation
  • Share your creations with #OpenSora

The future of video is open. Build it with us.

OpenSora: Where imagination becomes motion — freely, locally, and responsibly.

About

Open-source, locally runnable text-to-video generation — inspired by Sora, built for everyone. Run cinematic video synthesis offline, with full control and transparency.

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors