LTX MODELS

Upload a still image and turn it into a video. LTX reads the subjects, environment, and lighting from your image, then animates them based on your prompt.

How do you animate an image with LTX?

Upload your reference image and write a prompt that describes what happens next. Don't describe what's already in the image. Focus on motion, camera movement, and sound. LTX 2.3 improved image-to-video consistency, producing fewer frozen frames and less fake zoom compared to earlier versions.

Your image Works with concept art, product shots, storyboard frames, portraits, or any reference you want to bring to life. The model reads the composition and style from your image, so the output stays visually consistent with your starting frame.

Your prompt Describe the change, not the scene. If your image shows a woman on a balcony, don't write "a woman standing on a balcony." Write "the camera slowly pushes in as wind catches her hair, city lights flickering below, ambient evening noise."

The catch: Earlier LTX versions sometimes froze the image or added a slow pan instead of real motion. LTX 2.3 fixed this. If you're still seeing static outputs, add more specific motion cues to your prompt.

Transform existing footage into a new style while keeping the original motion and structure. Change the look, the lighting, the atmosphere, or enhance detail without reshooting.

How do you restyle video with LTX?

Upload your source video, write a prompt describing the target style or enhancement, and set the denoise strength. Lower denoise keeps more of your original footage. Higher denoise gives the model more freedom to restyle. The original motion and structure carry through either way.

Denoise / strength Want subtle enhancement and sharpening? Keep denoise low (0.2 to 0.4). Want a full style transfer that changes the look of every frame? Push it higher (0.6 to 0.8). The tradeoff: more style freedom means less fidelity to your original footage.

Your prompt Describe the target look. "Film noir, high contrast, deep shadows, desaturated with silver highlights" takes your footage in a specific direction. Be as visual and specific as you would with text-to-video prompts.