Approve an AI avatar video's first frame before the full render

Create a preview. Sume's avatar video previews generate only the first-frame still stage, so you can approve framing before spending a full Avatar Video generation. Regenerate the still until it is right, then call generate-video on the same preview.

The four calls

What a preview carries

• The create body matches Avatar Video: exactly one of script or video_inputs, plus optional product_image, scene, quality, aspect_ratio, title and captions.
• quality defaults to plus here when omitted.
• Multi-scene previews return one still per scene; later stills are pose-anchored continuations of the first frame when scenes share a background.
• Captions are never burned into preview stills. Caption intent stored on create applies only at generate-video.

When it is worth the extra step

Use a preview when the first frame decides the video: a product held in hand, a branded set, a multi-scene hook. Skip it for a repeat format where the framing is already proven; a direct call to the talking-video route is one request instead of three. The docs do not list a separate preview price in the pages I read, so check the live catalog rather than assume it is free.

Poll like any job

The create response includes job polling URLs. Use GET /v1/jobs/{id}/status then /result. Prefer resource_status for readiness and job_status for polling over the legacy status field.

Create a preview

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: preview-001" \
  -d '{"avatar_handle":"your_avatar","aspect_ratio":"9:16","script":"Three sentences that fit in about twenty seconds."}'