LTX-Video
A self-hosted AI video generation model with ComfyUI and Diffusers support, image animation, video editing, and fine-tuning tools.
LTX-Video is an AI video generation model for creators building controlled animations and developers adding video tools to their own products. You can run it locally or on your own servers using publicly available weights. The LTX family also offers a managed cloud API; local deployments can run in isolated environments without a cloud dependency.
It generates video from text or images, accepts multiple keyframes, and can extend footage forward or backward. Video-to-video editing lets you work with existing clips. Pose, depth, and Canny edge controls give you ways to guide movement and composition beyond a written prompt. ComfyUI and Hugging Face Diffusers integrations make it usable within existing generation workflows.
Model variants trade output quality against speed and GPU memory use. Distilled models suit repeated drafts, while smaller and quantized variants reduce VRAM needs. LTX-Video-Trainer supports full fine-tuning and LoRA training for adapting the model to a particular style or domain.
The newer LTX-2 and LTX-2.5 models are separate successors. They add synchronized audio and video; LTX-2.5 also supports connected shots and HDR and EXR output. The LTX-Video repository code uses Apache 2.0. Model licensing is separate: each checkpoint has its own terms. LTX-2.5 has conditional permissions based on organization size and separate commercial licensing.
