SaaS release note avatar video: 24 seconds, 3 scenes, cost on Sume

A 24 second release-note video with a Sume avatar costs $4.416 on standard, $5.88 on plus and $13.20 on max, with no product image. It fits the 4 to 60 second window and uses three ordered scenes in video_inputs: a spoken hook, a silent demo beat, and a spoken call to action. Write the script in English, since the clip prompt is English only.

The scene plan

Planning at roughly 2.8 words per second, the spoken parts should be short. The numbers below are a plan, and each scene needs an explicit duration that adds up to the total.

• Hook, 6 seconds: about 17 words naming the feature and the problem it removes.
• Demo beat, 10 seconds: voice.type: "silence" with a duration of 10, so the viewer looks at a cut-in screen recording you add later.
• Call to action, 8 seconds: about 22 words with the one action you want, such as turning the setting on.
• All three scenes share one background prompt, because the current execution expects one shared scene and one avatar per final video.

Review before you pay for the render

Release notes change at the last minute, so create an Avatar video preview first. It produces first-frame stills, one per scene, without starting the full render. If the stills are right, call generate-video on the preview id. You can set a different quality at that step without making a new preview, but a changed script or scene needs a new one.

Add inline captions if the video will play muted in a changelog page. Captions burn into the final MP4 at generation, and a caption failure is a soft failure: the job can still succeed with a clean video.

Where the demo footage goes

The silent beat is a hole, not a screen recording. Sume does not insert your product screen into it. Render the avatar clip, then lay your recording over the silent seconds in your editor or with the timeline tools in the models overview. If you only need a short spoken change summary and no demo, drop the middle scene and render 14 seconds: 14 x 0.184 = $2.576 on standard.

Request skeleton

The multi-scene body follows the documented shape. Each scene has an id, a voice block and a background block, and every scene shares the same background prompt. Send it with an Idempotency-Key such as the release version, so a retry for the same note returns the same job instead of creating a second paid one.

If you ship release notes every two weeks, 26 notes a year at 24 seconds on standard come to 26 x 4.416 = $114.82, or $152.88 on plus. That is the budget to compare with the cost of recording a person each time.