How to use the Retake endpoint

The Retake endpoint lets you replace a specific segment of an existing video without regenerating the whole thing. You define the start time and duration of the section to edit, and the LTX model regenerates just that portion.

What is Retake?

Retake is a targeted editing tool. Instead of starting from scratch when one part of a video doesn't work, you mark the segment you want to replace and the LTX model regenerates just that section — keeping everything else intact. You can replace the video, the audio, or both.

Retake is available on ltx-2-3-pro only.

LTX-2 models are being deprecated

If you're currently using ltx-2-pro for Retake, switch to ltx-2-3-pro before July 15, 2026. After that date, LTX-2 models are automatically served by LTX-2.3 at LTX-2 rates. On August 15, 2026, LTX-2 requests will return an error. See the LTX-2 deprecation notice for full details.

What you need to specify

Every Retake request requires three things:

  • The input video — the video you want to edit, passed as a URL or uploaded file
  • Start time — when the segment begins, in seconds
  • Duration — how long the segment is, in seconds (minimum 2 seconds)

Your input video must be at least around 3 seconds long (73 frames at 24 fps), and the maximum input resolution is 4K.

Optional settings worth knowing

  • Prompt — describe what should happen in the replaced segment. Without one, the API generates content based on the surrounding video context.
  • Mode — choose whether to replace video only, audio only, or both (the default is both)
  • Resolution — defaults to 1080p based on your input video's orientation if not specified. Note: Retake currently supports 1080p output only.

How is Retake billed?

Retake is billed per second of the input segment being replaced — not the full video length. So if you retake a 4-second segment of a 30-second video, you're billed for 4 seconds. See What does the LTX API cost? for current rates.

What if I get an error?

If a Retake request fails, the most likely cause is the input video rather than your code. Check that your video meets these requirements:

  • Format: MP4 or MOV
  • Codec: H.264 or H.265
  • Minimum length: ~3 seconds (73 frames at 24fps)
  • Maximum resolution: 4K (3840×2160)
  • Output resolution: 1080p only

If the same video works in the API Playground but fails through the API, that points to an endpoint-specific issue. Contact support and include your x-request-id response header, the endpoint, model, and a description of the input media.

For a full list of error codes, see API error codes explained.

Making your first Retake request

The Retake API reference has the complete parameter list, example requests, and response format.

 

 

We'd love your feedback. Did this article help?