Brand mascot as a reusable avatar: one photo, every ad video

To keep a brand character consistent across ads, create it once with POST /v1/avatar-1.0/generate and a photo input, then pass the same avatar_handle to every POST /v1/avatar-1.0/talking-video request. The handle is the identity. Each ad changes the script and scene, not the face.

Three ways to create the character

Avatar creation takes a top-level avatar_handle and an input union. You can describe the character in text (a prompt), give structured traits with the props type, or send a reference image with the photo type. Sume normalizes the handle and stores it without a leading @.

For a brand mascot or a founder, the photo route is the one that fixes the look. image_url must be a fetchable public HTTPS image. Before submission, Sume rejects localhost, private-network URLs, non-HTTPS URLs and non-image responses.

curl -X POST https://api.sume.com/v1/avatar-1.0/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: avatar-image-001" \
  -d '{
    "avatar_handle": "reference_presenter",
    "input": {
      "type": "photo",
      "image_url": "https://example.com/reference.png"
    }
  }'

Wait for the job, then list it

Each request creates a job. Poll GET /v1/jobs/{id}/status, then read /result when it is completed. After that, GET /v1/avatar-1.0/avatars lists your avatars and GET /v1/avatar-1.0/avatars/{id} reads one. Store the handle in your own config, not in each campaign.

Use the handle in every ad

The talking-video route uses the top-level avatar_handle to find a ready avatar. Provide one of script or video_inputs. Sume accepts a plan whose estimated duration is 4 to 60 seconds, so a longer story is several jobs joined later.

Because the current execution supports one resolved avatar for each final video, a two-character scene is two renders. Plan for that when you write a dialogue.

Preview before you spend

The docs point to avatar video previews for first-frame review before a full render. Use them when a new scene prompt changes the look. A consistent character is as much about the background and framing as it is about the face, so approve a preview whenever you change either.