Nodaro Docs
DocumentationNode ReferenceModelsAI Agents (MCP)DevelopersSelf-hostingResearch
Video

Retake Video

Regenerate a time window of an existing video with LTX 2.3 Pro. Replace only the sound, only the picture or both, and keep everything outside the window.

The Retake Video node regenerates a chosen time window of an existing video from a text prompt. You choose where the window starts, how long it lasts, and whether to replace the sound, the picture or both. Nodaro stitches the new content back into the clip without re-rendering the frames outside the window, so the result is as long as the source.

When to use it

  • A few seconds of dialogue need a different line, and the picture is fine.
  • One shot looks wrong, and the soundtrack must stay in sync.
  • A short moment needs a complete reshoot, picture and sound.
  • One beat of a clip needs a new sound effect or a different score.

For a change of 4 seconds or more that needs image references, use Edit Video Pro. To change a whole clip, use Video to Video.

Quick start

Add the node

Press Tab on the canvas and choose Video › Continue & Restyle › Retake Video.

Connect the clip

Wire the source video into the Video input. The node reads the clip's length and sizes its scrubber to it.

Choose the window

Drag the two handles of the scrubber on the node. The first handle is the start of the window, and the second is its end. You can also type Start time (s) and Duration (s, min 2) in the settings panel. The window must be at least 2 seconds long.

Choose the mode and write the prompt

In the node toolbar, set Mode to Replace audio, Replace video or Replace both. Then describe what should happen in the window. The prompt applies only to the selected window.

Run it

Click Run on the node. The result is the full clip with the window replaced.

videopromptlookGenerate Video10-second dialogue sceneTextHe says: see you tomorrowCamera MotionSlow push-inRetake VideoReplace both, 4 s to 7 s
Retake Video reshoots seconds 4 to 7 of a generated scene with a new line of dialogue and a slow push-in, and keeps the rest of the clip.

Inputs

InputAcceptsWhat it does
VideoVideo nodes, such as Generate Video, Upload Video, Video to Video and Extend VideoThe source clip. Required.
PromptText nodes and pickersWhat the new content should be. A wired prompt replaces the prompt typed in the settings panel.
LookLook pickers, such as Setting, Lens, Lighting, Mood, Style, Color / Look and Camera MotionEach picker adds its choice to the prompt. A Camera Motion picker also sets the model's own camera motion when its choice has a match.

The output, Video, is a full-length clip with the same duration as the source. Only the selected window is new.

Settings

The node toolbar shows the settings you change most. The settings panel shows all of them.

SettingWhat it does
ModelLTX 2.3 Pro, the only retake model.
ModeReplace audio, Replace video or Replace both (the default). In the settings panel, the same choice is called Retake Mode.
Aspect Ratio16:9 or 9:16. The default is the aspect ratio of the source video. A different ratio letterboxes or crops the window only.
VersionsHow many versions to generate in one run, from 1 to 4. Each version is billed separately.
PromptWhat should change in the selected window.
Start time (s)Where the window starts. The default is 0. The first scrubber handle and this field always match.
Duration (s, min 2)How long the window lasts. The default and the minimum are 2 seconds.
ResolutionLocked to 1080p for retakes.
FPS24, 25, 48 or 50 frames per second.
Generate audioOn by default. When it is off, the new picture has no generated sound. The setting has no effect in Replace audio mode.
Pre & post textText that is always added before and after the prompt. It is hidden from people who use your workflow as an app. See Prompt pre and post text.

The three modes

ModeKeepsRegeneratesUse it to
Replace audioThe original pictureThe sound in the windowSwap a line of dialogue, add a sound effect or re-score a beat
Replace videoThe original soundThe picture in the windowRepaint a shot while the soundtrack stays in sync
Replace bothNothing in the windowPicture and soundReshoot the moment

Model

Retake Video runs on LTX 2.3 Pro. LTX 2.3 Fast has no retake mode.

ModelMakerModesCreditsDetails
LTX 2.3 ProLightricksImage to video, Text to videofrom 40Lightricks LTX 2.3 Pro — text/image/audio→video up to 4K, 6/8/10s, end-frame interpolation.

Credits

Retake Video costs 40 credits per second of the window, at 1080p, which is the only resolution:

credits = 40 × window seconds × versions

WindowVersionsCredits
2 seconds180
4 seconds1160
6 seconds1240
10 seconds1400
4 seconds3480

Extend Video uses the same rate when it runs on LTX 2.3 Pro.

Limits

  • The window is rendered at 1080p. If the source is larger than 1080p, the window is 1080p and the untouched parts keep the source resolution. For a uniform result, run the video through Transcode Video afterwards.
  • The window is at least 2 seconds. The scrubber enforces it, and a shorter window is refused before the run.
  • There is no image input. The retake model takes the prompt and the look pickers only. Describe the person or the object in the prompt, and wire a picker such as Style into Look to keep a style.

Tips

  • Choose the smallest mode that fixes the problem. Replace audio keeps the picture exactly as it was, and Replace video keeps the sound in sync.
  • Generate versions, then choose. Set Versions to 3 or 4 and keep the best take. Each version costs the same as one run.
  • Cut without regenerating. To remove a part of a clip rather than replace it, use Trim Video.

Frequently asked questions

Last updated on

On this page