Passer à la navigation
Guides de l’app web

Speaker Selection — Studio

Speaker selection lets you pick which person gets lipsynced in a video or image with multiple people. You can manually click a face, or for videos, use Active Speaker Detection to let Sync Labs identify the speaker automatically. If a detection lands on the wrong person, drag or resize its box onto the correct face. If no face is detected, you can place a box yourself. For programmatic usage via the API, see the API guide.

Speaker selection is available with our lipsync-2, lipsync-2-pro, and sync-3 models in both Lite and Advanced modes. Note that react-1 does not currently support this feature. For sync-3 image inputs, only manual face selection is supported — Active Speaker Detection is video-only.

Selecting a Speaker in Video

1

Upload your video

Upload a video that contains multiple people. Sync Labs detects the number of speakers in the background.

Studio overview

2

Enable face detection

Click the icon in the video player controls. Green bounding boxes appear around detected faces with a hint: “select which speaker to lipsync.” If no boxes appear, see When No Face Is Detected below.

Face detection active

3

Click on the speaker's face

Click on the bounding box of the person you want to lipsync. The selected face gets a bright green border and a face thumbnail appears in the controls.

Face selected

4

Generate

Click the Sync Labs button. The speaker configuration is sent automatically with your generation request.

Changing or Clearing Your Selection

  • Click the X on the face thumbnail to clear your selection.
  • Scrubbing through the video re-runs detection.
  • Click another face to switch speakers.
  • If the selected or auto-detected box lands on the wrong person, drag the box onto the correct face, or drag a corner to resize it — the face thumbnail updates to match.

When No Face Is Detected

When no face is detected in your video, the hint offers a place a box action. Choosing it places a box on the current frame. Drag or resize it onto the intended face, the same way you’d correct a wrong detection.

Active Speaker Detection

As an alternative to manual selection, toggle Active Speaker Detection in the Studio settings panel. Sync Labs identifies the speaker via lip movement analysis — no manual click needed.

Manual selection and Active Speaker Detection are mutually exclusive — you can only use one at a time. Active Speaker Detection may not work reliably on silent or low-motion clips.

Selecting a Speaker in Images (sync-3)

For images with multiple faces, you can select which person to lipsync using the same face detection overlay.

1

Upload your image

Upload an image with the sync-3 model selected. Face detection runs automatically when the image loads.

2

Click on the speaker's face

If multiple faces are detected, click on the face you want to lipsync. The selected face gets a bright green border and appears in the controls.

3

Generate

Click the Sync Labs button. The speaker coordinates are sent with your generation request.

Active Speaker Detection is not available for image inputs — only manual face selection. If your image contains a single face, the system skips the selection step automatically.

  • Lipsync Model — learn about supported models including lipsync-2 and lipsync-2-pro
  • Quickstart — get started with your first Sync Labs generation