v1.1.0

jitsucom/jitsuv1.1.0May 7, 2024by FrankLeeeee

AI Summary

Open-Sora 1.1 is released, featuring model weights for various resolutions and a new automatic data processing pipeline, along with architectural improvements to the ST-DiT model.

Key Highlights

Model weights available for 0s-15s, 144p-720p videos
Automatic data processing pipeline with scene cutting and filtering
Improved ST-DiT architecture with rope positional encoding and qk norm
Support for training with any resolution, aspect ratio, and duration
Support for image and video conditioning, editing, and animation

New Features

ST-DiT architectural improvements
Automatic data processing pipeline
Flexible resolution and duration support
Video editing capabilities
Image to video animation

Full Release Notes

📍 Open-Sora 1.1 released
- 🌠 Model weights are available [here](https://github.com/hpcaitech/Open-Sora). It is trained on 0s~15s, 144p to 720p, various aspect ratios videos. See our [report 1.1](https://github.com/hpcaitech/docs/report_02.md) for more discussions.
- 🔧 Data processing pipeline v1.1 is released. An automatic [processing pipeline](https://github.com/hpcaitech/Open-Sora#data-processing) from raw videos to (text, video clip) pairs is provided, including scene cutting, filtering(aesthetic, optical flow, OCR, etc.), captioning managing. With this tool, you can easily build your video dataset.
- ✅ Improved ST-DiT architecture includes rope positional encoding, qk norm, longer text length, etc.
- ✅ Support training with any resolution, aspect ratio, and duration (including images).
- ✅ Support image and video conditioning and video editing, and thus support animating images, connecting videos, etc.

Visit the [Open-Sora Gallery](https://hpcaitech.github.io/Open-Sora/) to view more samples.