Add a silent beat to an AI avatar video with voice type silence

To put a pause in a Sume avatar video, add a scene to video_inputs with voice type silence and a duration in seconds. The scene has no speech. Duration is required, and script or input_text are not permitted on a silence scene. Spoken scenes keep using type text with exactly one of script or input_text.

The scene fields

From Generate avatar video, read 2026-10-08:

Where a silent beat helps

A beat of quiet is useful when something on screen should carry the moment: a product being shown, a number appearing, or a caption landing. It also gives a hook room. A UGC-style clip might open with a spoken question, hold for a few seconds on the demo, then close with a call to action.

• Hook: spoken, 2-4 seconds.
• Demo: silence, enough time to show the product.
• Call to action: spoken, the final line.

Constraints that still apply

Silence counts toward the total. The planned duration of all scenes together must stay in the 4-60 second window. The execution supports one resolved avatar per final video, and it expects every scene background to resolve to one shared scene, so a silence beat does not give you a new location.

Preview the result before a full render. Preview creates one first-frame still per scene, and later stills are pose-anchored continuations of the first frame. Because a silence scene still gets a still, you can check that the avatar holds a natural pose through the pause. Details are on Avatar video previews.

Example body

Three scenes with a silent middle, shortened from the docs example.

{
  "avatar_handle": "sume_clawra",
  "aspect_ratio": "9:16",
  "video_inputs": [
    {"id": "hook",
     "voice": {"type": "text", "script": "Wait, one photo became a video?", "duration": 3},
     "background": {"type": "prompt", "prompt": "Casual bedroom framing"}},
    {"id": "demo",
     "voice": {"type": "silence", "duration": 4},
     "background": {"type": "prompt", "prompt": "Casual bedroom framing"}}
  ]
}

Timing a pause

A silence scene is plain time. Four seconds feels long on camera, so use it only when something visibly happens: a product turning, a result appearing, or a step on a screen. For a pure beat of emphasis, one to two seconds is usually enough, but remember the clip's total must be at least four seconds.

Because every scene shares one background, a pause cannot move the speaker to another location. If you need a different setting, make a separate video and join the clips afterward. This keeps the avatar's identity and the scene consistent inside each render.

Captions follow the spoken text only, so silence scenes carry no caption words. Viewers will see a gap, which is usually what you want, but check the preview and the final result on a phone before you publish.