Tutor intro video: approve the first frame, then pick the tier

To make a tutor intro video you can check before you pay for it, create an avatar video preview, review the first-frame still, and call generate-video on the preview id. The preview stills are independent of the quality tier, so you can approve once and choose standard, plus or max for the final render without making a new preview.

The flow is from Avatar video previews and Generate avatar video, read 2026-10-04.

One avatar, many intros

Avatar 1.0 is two steps: create a reusable avatar, then use its handle to make talking videos. A tutor can build the avatar from a photo with input.type: "photo", a prompt, or structured traits, and keep the handle for every later intro.

Create the preview

The create body matches the Avatar Video fields: exactly one of script or video_inputs, plus optional scene, quality, aspect_ratio, title and captions. Captions you set here are stored and burned in only when you generate the video, because preview stills are never caption-burned.

curl -X POST https://api.sume.com/v1/avatar-video-previews \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: tutor-intro-preview-001" \
  -d '{
    "avatar_handle": "math_tutor",
    "script": "Hi, I am Ms. Rivera. In our first session we will find where algebra stops making sense, and fix that first.",
    "scene": { "type": "prompt", "prompt": "A bright home study with a whiteboard behind the tutor" },
    "aspect_ratio": "16:9",
    "quality": "standard",
    "captions": { "enabled": true, "style": "slam", "language": "auto" }
  }'

Review, regenerate, generate

The response includes job polling URLs and an avatar_video_preview_id. When the preview is ready, GET /v1/avatar-video-previews/:id returns preview_image_url, and scene_previews[] for multi-scene plans. If the framing is wrong, POST /v1/avatar-video-previews/:id/regenerate reuses the stored request and refreshes only the stills.

The limits that still apply

Final render: curl -X POST https://api.sume.com/v1/avatar-video-previews/$PREVIEW_ID/generate-video with a body of {"quality": "max"}, or an empty body to keep the tier chosen at create. Admission, the ledger reservation and the provider submit all use that effective tier.

Structural fields (script, video_inputs, avatar_handle, scene, aspect_ratio) cannot be overridden at this step. Changing the script means a new preview. The video must still estimate to 4 to 60 seconds.