LTX 2 Retake
Re-generate audio, video, or both from an existing clip. LTX 2 Retake lets you improve, refresh, or change specific elements of a video you've already generated — without starting over.
What is LTX 2 Retake?
LTX 2 Retake is a unique model in the LTX family — instead of generating new video from scratch, it takes an existing video clip as input and re-generates specific components of it. You choose whether to retake the audio track, the visual content, or both.
Retaking just the audio means the visual motion and composition stay exactly as they are, but the audio track is regenerated — useful when the audio from an initial generation didn't match the visual mood, or when you want to try different sound character for the same clip.
Retaking the video while keeping audio locked allows you to change the visual style or content of a clip while preserving an audio track you're happy with. Retaking both gives you a fresh generation seeded from the structure of the original, producing variations that stay within a similar creative territory.
Re-generate audio
Replace audio, keep visuals
Re-generate video
Replace visuals, keep audio
Re-generate both
Fresh take from the same base
Clip refinement
Improve existing generations
How to use LTX 2 Retake on project.video
Open the composer
Go to your project.video dashboard. LTX 2 Retake is available in the model selector under Lightricks models.
Select LTX 2 Retake
Choose LTX 2 Retake. The composer will show an input slot for your existing video clip and options for what to retake.
Upload your clip
Upload the video clip you want to retake. This becomes the base for the re-generation — the model uses its structure to guide what it produces.
Choose what to retake
Select whether to retake audio only, video only, or both. Update your prompt if needed to guide the new generation, then generate the retake.
Technical specs
Best use cases
Audio that doesn't match the visual mood
Generated the perfect visual clip but the audio missed the tone? Retake audio only to get a fresh sound pass while keeping every frame of the video exactly as generated.
Visual refresh of a working audio track
Have an AI-generated or sourced audio track that's perfect but visuals need work? Lock the audio and retake the video to find visuals that match the audio character.
Variation generation from a strong base
Found a strong base clip but want to see alternatives? Retake both components to generate variations that share the structural DNA of the original without being identical.
Production quality improvement
Retake a clip with a more detailed prompt to improve its output quality. Use the original as a reference while the model generates a fresh version guided by your refined instructions.