🎬 LTX 2.3 AI Video Model Guide

What is LTX 2.3?

LTX 2.3 is the latest version of Lightricks' open-source AI video generation model. It offers improved visual quality, better prompt understanding, and enhanced text rendering compared to previous versions.

Quick Start (ComfyUI)

1. Install ComfyUI if you haven't already: comfy.org/download
2. Install LTXVideo nodes via ComfyUI Manager → Custom Nodes → search "LTXVideo"
3. Download models from HuggingFace (see below)
4. Load workflow from Comfy Manager → LTX-2.3 workflows

Model Variants

• Full Dev (bf16) - Best quality, requires high VRAM
• Quantized (fp8) - Lower VRAM usage with good quality
• Distilled - Faster inference, good for quick testing
• Spatial/Temporal Upscalers - For enhancing output quality

Useful Links

📚 Documentation

• LTX Documentation - ComfyUI Integration
• Comfy.org LTX 2.3 Workflow Examples
• GitHub - ComfyUI-LTXVideo

🤗 Models

• LTX-Video on HuggingFace

📹 Tutorials

• Run LTX 2.3 Video Generation AI Model Locally with ComfyUI
• Install LTX-2 in ComfyUI on Windows

Features

• Text-to-Video - Generate videos from text prompts with improved prompt understanding
• Image-to-Video - Animate input images with consistent motion
• Portrait Video - Specialized mode for talking head / portrait content
• Text Rendering - Better text overlay capabilities in generated videos

Tips

• Save output locally first (e.g., ComfyUI/output/) - avoid syncing folders like OneDrive
• Use the quantized (fp8) model if you have limited VRAM
• ComfyUI Manager has pre-built LTX-2.3 workflows to get started quickly