Turns a text prompt or a single image into a short video

You type a scene description, or drop in a single image, and it generates a short video clip. Both modes are native: the same model handles text-to-video and image-to-video, so you can animate an existing photo without switching to a different tool.

It is a community-trained model built on Lightricks' LTX 2.3 video model, and the project ships the practical pieces around it: ready-made ComfyUI workflows, a speed-up add-on that cuts the number of rendering steps so clips finish faster, and a local prompt helper you can load in LM Studio that expands a one-line idea, optionally together with your image, into the detailed prompt the model responds to best.

There are two builds: a smaller one for graphics cards with less memory and a full-quality one. With over a million downloads it is currently one of the most-used community video releases on Hugging Face.

There is no browser demo, so this one is for people willing to run ComfyUI locally. Download one of the dev builds plus the speed-up add-on from the model page and load the provided workflow; the author notes you should use the add-on or the full model, not both at once. For a first test, animate one photo with a short motion prompt, then generate the same scene from text alone and compare which route gets closer to what you pictured. A recent consumer GPU is required.

Documentation is thin: setup notes are sparse, official inference code is promised but not shipped yet, and you should expect some trial and error in ComfyUI. The model also has no built-in content restrictions, so what comes out is entirely down to your prompts.

• Animate a still product shot into a short clip for a landing page.
• Turn concept art or an illustration into a moving scene.
• Rough out camera moves for a scene before filming anything.
• Generate short background footage you cannot easily film yourself.

Try the demo

View model page

Read this issue on neuralexpedition.com →