Skip to content
NewsLightricks Releases Real-Time Video Generation Model LTX-Video

Lightricks Releases Real-Time Video Generation Model LTX-Video

LTX-Video Demo

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.