v1.3

sherlock-project/sherlockv1.3Feb 21, 2025by zhengzangw

AI Summary

Open-Sora V1.3 introduces improvements in model architecture and generation quality, while expanding support for image-to-video generation and video extension with higher resolution and frame lengths.

Key Highlights

  • Upgraded Video Compression
  • Enhanced ST-DiT Architecture
  • More High-Quality Data & Multi-Stage Training
  • Advanced Model Conditioning and Generation Quality

New Features

  • Image-to-video generation
  • Video extension
  • Expanded resolution support
  • Expanded frame length support

Full Release Notes

We are excited to introduce Open-Sora V1.3.0, bringing improvements in model architecture and generation quality.

Open-Sora V1.3.0 Release šŸŽ‰

### šŸš€ Key Highlights:
šŸ“¦ Upgraded Video Compression
šŸŒ€ Enhanced ST-DiT Architecture
šŸŽžļø More High-Quality Data & Multi-Stage Training
✨ Advanced Model Conditioning and Generation Quality

Open-Sora 1.3 continues to support image-to-video generation and video extension with expanded resolution and frame length support.

šŸ”— Check out our [technical report](https://github.com/hpcaitech/Open-Sora/blob/main/docs/report_04.md) and [gallery](https://hpcaitech.github.io/Open-Sora/) to explore more!