Avatar photo scene: a talking host in your showroom for year-end

Scene as photo, scene as prompt

The avatar-video request accepts scene: { "type": "photo", "image_url": "https://..." } for a photo scene reference, or scene: { "type": "prompt", "prompt": "..." } for text direction. For a year-end sale in a real space, use the photo form with a picture of the showroom, shop floor or counter.

Media fields must be fetchable public HTTPS URLs. Localhost, private-network and non-HTTPS URLs are rejected before generation.

A single-script request

Omit product_image for a productless clip, or pass a product image to feature one item. Use script for a single speaker; the estimated duration must be 4 to 60 seconds, so shorten longer scripts or split them into multiple jobs.

curl -sS -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: showroom-yearend-001" \
  -d '{
    "avatar_handle": "sume_clawra",
    "script": "Year-end clearance starts today. Floor models are marked down until the thirty-first.",
    "scene": { "type": "photo", "image_url": "https://cdn.example.com/showroom.jpg" },
    "aspect_ratio": "9:16",
    "quality": "plus"
  }'

Choose the photo for the shot

Because the avatar is placed into the scene, choose a photo with clear floor space and even light. Use one photo per clip; the current execution supports one resolved avatar per final video and expects scene backgrounds to resolve to one shared scene.

When the location changes

A clip that needs two locations is two jobs joined afterwards, since one request resolves to one shared scene. Generate each clip with its own scene, then assemble them with a Sume timeline route. Keep the same avatar_handle so the host stays the same person.