🗣️ How do I lip-sync only one character when there's more than one in a scene?

Atlabs videos can have as many characters as you need, each with their own voice. Every character can speak in any scene. Tag who’s speaking so the right mouth moves.

If a scene has two or more characters and you want only one speaking, Atlabs needs to know exactly who; otherwise, every character in frame can end up mouthing along to the same dialogue. It lives in the Lip Sync tab, inside Edit Video for that scene.

Tag the speaking character

1. Open the scene and go to Edit Video → Lip Sync tab.
2. Select OmniHuman v 1.5 Model from the drop-down.
3. In your prompt, tag the character who should be speaking with @, then describe the action, for example: "@Lumi Fox is speaking the dialogue." Make sure the character has been added to the story.
4. Leave the other characters in the scene untagged so they aren't read as speakers.
5. Generate. Only the tagged character's mouth moves to match the voiceover; everyone else stays still.

Use the selected scene audio

Before generating lip sync, open the scene's Audio tab and verify which audio track is selected and who is speaking. In the Lip Sync tab, apply lip sync to that scene and tag the intended speaker with @ in the prompt. Make sure the character has been added to the story. Generate lip sync for one scene first and check the result before applying it to other scenes.

If the output uses newly generated audio instead of the selected track, try a different lip-sync model; OmniHuman v1.5 is one option to try if MiniMax does not work as expected.

Reapply lip sync after changing dialogue

If you change or regenerate a line in a shot, reapply lip sync to that shot so the mouth movement matches the updated audio. In a multi-character shot, use the prompt to tag the specific character who should speak. You can fix one shot without regenerating the rest of the video.

💡 Tip: The more specific your tag, the more reliably it works; adding what the speaking character is wearing or doing alongside the @ tag helps Atlabs lock onto the right one when a scene is busy.

Multiple characters speaking at the same time

To have multiple characters speak simultaneously in one scene, use a Video motion model, such as MiniMax H3 Reference to Video. Add the scene image and multiple audio tracks as references in the prompt.

Updated on: 10/16/2026