AI spokesperson release notes, 90 seconds: two jobs, cost by tier
A 90-second spokesperson video is two Sume avatar jobs because one job tops out at 60 seconds. Cost at standard, plus and max, and how to split the script.

A 90-second release notes video from an AI spokesperson takes two Avatar 1.0 jobs, because Sume accepts a target duration of 4 to 60 seconds for each job. Split the script into two parts of 45 seconds, render both, and join them. At the plus no-product rate that is $22.05 for the 90 seconds of avatar video. Sume avatars are async jobs: you submit a request, get a job id back, and read the finished video later. There is no live video session.
This post gives the cost at all three tiers and a way to split a script without the join showing.
Cost of 90 seconds
| Tier | Rate per second | One 45 s job | Both jobs (90 s) |
|---|---|---|---|
| standard | $0.184 | $8.28 | $16.56 |
| plus | $0.245 | $11.025 | $22.05 |
| max | $0.55 | $24.75 | $49.50 |
Splitting the script
Cut at a natural break, such as between "what is new" and "what is fixed", so the join reads as a change of topic and not as a jump in a sentence. Write each half to land near 45 seconds. The docs say Sume estimates the duration from the script, and rejects a script it estimates above 60 seconds, so leave margin: 45 seconds each is comfortable, and it also leaves room if a part runs longer than you guessed.
Use the same avatar_handle, the same aspect_ratio and the same quality for both jobs so the face, framing and finish match. If you want a recognizable scene, pass the same scene for both. The current execution supports one resolved avatar per final video, which is another reason to keep one presenter across the two parts.
Joining the parts
Sume's Timeline route can compose the final video from the two results. The fixed-price table in provider-pricing lists a Timeline render at $0.10 per output minute, rounded up, so a 90-second output is two billable minutes, or $0.20. Add that to the avatar cost if you let Sume join the clips: at plus, $22.25 in total. If you join the clips in your own editor, the second line does not apply.
If you need burned-in captions, you can request them on each avatar job with the captions field. Inline captions do not create a separate billed caption job.
Run the two jobs
Submit both with different idempotency keys, poll both, and join when both are completed. If one fails, retry only that part.
for part in 1 2; do
curl -s -X POST https://api.sume.com/v1/avatar-1.0/talking-video \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: release-notes-2026-10-part-$part" \
-d "$(cat part-$part.json)"
doneWeekly cadence cost
If you publish release notes every week, the annual figure matters more than the single clip. At standard with no product image, a 90-second video is $16.56, so 52 of them cost $861.12 over a year before any Timeline joins. At plus it is $1146.60, and at max it is $2574.00. Those numbers come from the same per-second rates, so they change only if Sume changes the rate card.
A common pattern is standard for the weekly update, where speed and cost matter, and max only for a quarterly roundup that you will keep on your site. Because the tier is a request field, you can switch per video without creating a new avatar.
Review before you pay for the full render
For a 90-second video, an approval step is worth having. An Avatar video preview generates the first-frame still for each scene, and you can regenerate the stills until the framing is right. Then generate-video renders from the preview, and an optional quality there overrides only the final render tier. Preview stills are tier-independent, so approving at one tier and rendering at another does not need a new preview.
Write release notes as a list of facts first and a script second. Each sentence should name one change and one benefit, in plain words, because a spokesperson reads every word aloud and a viewer cannot skim. A script that sounds fine on a page can sound dense when spoken, so read it aloud once before you submit. A paid render is a poor place to find a tongue-twister, and two 45-second jobs mean any fix costs a re-render of a whole half.
Sources
Related posts
More in Use cases
- How do I make a spoken wake-up alarm with an AI voice and music?
A wake-up alarm is a short voice greeting joined to a music file: 2 cents of TTS and a 1-cent concat per day, with one $0.125 music track reused all week.
- Animate a painting with AI video: first frame and a slow camera prompt
Pin a painting as the first frame with frame_images on Sume, ask for a slow camera move, and pick the model by length and price. A 6-second price table.
- Animate a pencil sketch or storyboard panel as the first frame
Use a sketch or storyboard panel as first_frame on Sume, and the next panel as last_frame on models that take an end frame. Prompt, models and windows.
- How do I animate my podcast cover art into a 5-second video intro?
Use the cover as a first frame for a 5-second clip ($0.63 at 720p), add a $0.125 sting and a $0.10 render: $0.855 once, reused every episode.
Written by Sume