Lightricks Releases Real-Time Video Generation Model LTX-Video
Lightricks has recently launched a groundbreaking AI video generation model, LTX-Video. This is the first real-time video generation model based on DiT (Diffusion Transformers), capable of producing high-quality videos at a rate faster than playback speed.
Key Features
- Real-time generation: Capable of producing videos at 24FPS and 768x512 resolution, with generation speed faster than playback speed
- Supports both text-to-video and image-to-video generation modes
- Trained on a large-scale video dataset, ensuring realistic and natural content generation
- Open-source implementation, supporting local deployment and ComfyUI integration
Technical Details
The model supports the following features:
- Resolution must be a multiple of 32
- Frame count must be 8n+1 (e.g., 257 frames)
- Optimal working resolution is below 720x1280
- Recommended frame count is below 257 frames
- Supports detailed English prompt descriptions
Open Source and Deployment
LTX-Video is open-sourced on GitHub and Hugging Face:
Developers can use it through the following methods:
- Online Demo
- ComfyUI Plugin
- Local Deployment
This groundbreaking real-time video generation technology opens up new possibilities for creative professionals and content creation. Whether for movie previews, advertising production, or social media content creation, LTX-Video enables rapid video creative expression.