Sign in

v1

latestOpenAPI 3.1.02026-08-04143345.6 KB
asyncVideoGeneration

Retake video section

Re-generate a section of an existing video by replacing either its audio, its video, or both. Select a time range and mode to iterate on part of a generation without starting over — the rest of the video is preserved.

Responds immediately with the job id and created_at timestamp. Poll GET /v2/retake/{id} until the status is completed, then download from result.video_url.

Billed per second of the input video. See Pricing.

post/v2/retake

Headers

Authorizationstring required

API key authentication

Request body

video_uristring required

Input video for editing. See Input Formats for supported formats and codecs.

  • Maximum resolution: 3840x2160 (4K)
  • Minimum frame count: 73 (around 3 seconds at 24fps)
promptstring

Describing what needs to happen in the generated video in a section defined by start_time and duration.

start_timenumber double required

Start time in seconds, defines the section to be edited in the input video.

The section defined by start_time and duration should be within the duration of the input video. If the section extends beyond the duration of the video, it will be clamped to the video duration.

durationnumber double required

Duration in seconds, defines the duration of the section to be edited in the input video. Must be at least 2 seconds.

The section defined by start_time and duration should be within the duration of the input video. If the section extends beyond the duration of the video, it will be clamped to the video duration.

mode'replace_audio' | 'replace_video' | 'replace_audio_and_video'

Can be any of the following:

  • replace_audio - Replace only the audio track
  • replace_video - Replace only the video track
  • replace_audio_and_video - Replace both audio and video tracks

The default value is replace_audio_and_video if not specified.

resolution'1920x1080' | '1080x1920'

The resolution of the generated video in WIDTHxHEIGHT format. When omitted, the resolution is automatically determined based on the input video orientation — portrait videos produce 1080x1920 output, landscape videos produce 1920x1080 output.

model'ltx-2-pro' | 'ltx-2-3-pro'

Model to use for video generation.

Deprecated: ltx-2-pro is deprecated. See the LTX-2 Deprecation Notice.

Response

Job submitted successfully

idstring required

Unique job identifier. Use this to poll for status.

created_atstring date-time required

ISO 8601 timestamp of when the job was created.