One wrong sentence in an avatar script: what 5 re-renders cost

If you spot a wrong sentence in an avatar script, you fix the script and run a new job; five full re-renders of a 20-second take cost $18.40 on standard, $24.50 on plus and $55.00 on max. The preview docs say structural fields such as script, video_inputs, avatar_handle, scene and aspect_ratio need a new preview when they change (Avatar video previews).

Revision budget

Each round is a fresh render at the tier rate, with no product image.

What does not need a new preview

Preview stills are tier-independent. After you approve the stills you can call generate-video on the preview id with a different quality, and Sume reuses the first frame. Changing tier alone therefore never needs another preview.

• Wording, scenes, avatar, ratio changed: new preview, then a new final.
• Only the tier changed: call generate-video on the same preview with a new quality.
• Captions stored on preview create are applied at generate-video, never burned into stills.

Draft cheap, finish once

Iterate the wording on standard at $0.184 per second, then render the approved script once on max at $0.55. Four standard drafts plus one max final of 20 seconds costs $25.72, against $55.00 for five max renders.

Guard rail

Use a stable Idempotency-Key per script revision, not per attempt, so a network retry of revision 3 cannot bill twice.

# revision 3 of the Black Friday script
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: bf-script-rev-3" \
  -d '{"avatar_handle":"product_host","quality":"standard","script":"Revised wording goes here."}'