Avatar spokesperson in front of your shop photo: scene type photo

Yes. The Sume Avatar 1.0 talking-video route takes a scene object, and scene: { "type": "photo", "image_url": "https://..." } uses your photo as a photo scene reference. That is the way to put a spokesperson in front of your own shop, counter or kitchen rather than in a prompted setting.

It is a photo scene reference, not a guarantee of a pasted background, so a result can differ from your shop. Check the first frame before the full render, and do not use it where the exact look of the premises is a legal or safety claim.

Prompt scene or photo scene

The two forms of scene are in the Sume docs. Pick by what you need to control.

Request

Media fields must be public HTTPS URLs that Sume can fetch. Output is 720p. The default aspect ratio is 9:16, and 1:1, 3:4, 4:3 and 16:9 are also accepted.

curl -X POST https://api.sume.com/v1/avatar-1.0/talking-video \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: avatar-shop-photo-001" \
  -d '{
    "avatar_handle": "store_host",
    "aspect_ratio": "9:16",
    "quality": "plus",
    "scene": { "type": "photo", "image_url": "https://example.com/shop-counter.jpg" },
    "script": "Welcome to our shop. This Saturday we are open late for Small Business Saturday. Come and see what is new."
  }'

Checks before posting

Run the preview route first: /v1/avatar-video-previews returns a first frame, then generate-video finishes the job. Look at the shop details, the signage and any visible text, since generated text in a scene can be wrong.

With video_inputs, the scene backgrounds must resolve to one shared scene, and the current run uses one avatar per final video. So for a multi-scene plan use the same photo in every scene.

• Take the photo in good light and without people in it.
• Do not use a photo of a place that is not yours.
• Keep the date and hours in the script short, and double-check them.