Weekly recap reel from six clips: what 52 Fridays cost on Sume
A 60-second weekly recap from six clips costs $0.42 in Sume media jobs, or $21.84 over 52 weeks. A music bed adds $0.125 once; Agent turns bill separately.

A weekly recap reel built from six phone clips, rendered as one 60-second vertical video with burned-in captions, costs $0.42 per week in Sume media jobs: six trims at $0.02, one timeline render at $0.10 and one caption job at $0.20. Over 52 Fridays that is $21.84, plus a one-time $0.125 if you want a music bed generated once and reused.
That number only covers the deterministic media steps. If a scheduled Agent decides which clips to use, its own LLM turn is billed on the Agent wallet and is not part of the figures below. The rates here come from the Sume model pages for each tool, read on 2026-10-03; the docs tell you to confirm them live in GET /v1/catalog before you budget a year.
The weekly job list
Order matters because each step takes a Sume-hosted URL from the one before it. Trim each raw clip to the moment you want, assemble the trimmed clips on one timeline with a short fade between them, then burn captions last, because the caption job reads a finished video and cannot be undone by a later render.
Keep the finished recap at 60 seconds or less. The caption page prices a standalone caption job at $0.20 for videos up to 60 seconds under the current fixed estimate, and says nothing fixed about longer ones. The timeline bills $0.10 for each output minute rounded up, so a 61-second recap would already bill as two render minutes.
| Step | Endpoint | Rate | Count | Subtotal |
|---|---|---|---|---|
| Trim each clip | POST /v1/video-trim | $0.02 per job | 6 | $0.12 |
| Assemble and fade | POST /v1/timeline-1.0/render | $0.10 per output minute, rounded up | 1 (60 s) | $0.10 |
| Burn captions | POST /v1/video-captions | $0.20 per job up to 60 s | 1 | $0.20 |
| Per week | $0.42 | |||
| 52 weeks | $21.84 |
What changes with the clip count
Only the trim line scales with the number of clips. Adding a clip adds $0.02, and 52 weeks of one more clip is $1.04. The render and caption lines stay flat as long as the video stays inside one minute.
A music bed is the one item you should not repeat weekly. Music 1.0 is a fixed $0.125 per accepted generation, and the timeline's soundtrack object takes a URL with loop and duck_db options, so one generated bed can sit under every episode. Ducking needs a real audio spine, not a silent one, so the bed only ducks under speech that your clips carry.
| Clips per recap | Trims | Weekly total | 52 weeks |
|---|---|---|---|
| 4 | $0.08 | $0.38 | $19.76 |
| 6 | $0.12 | $0.42 | $21.84 |
| 8 | $0.16 | $0.46 | $23.92 |
| 10 | $0.20 | $0.50 | $26.00 |
Who runs it on Friday
You have three ways to fire this each week, and they differ in what you pay beyond the table above. Calling the five endpoints from your own cron costs exactly the table and uses no Agent turn, but you write the clip selection yourself. A Scheduled run is a saved Agent automation on a 5-field cron expression in an IANA timezone, for example 0 16 * * 5 for Fridays at 16:00 in the timezone you pick; it is created in the Agents dashboard, because the Developer API can list and start schedules but cannot create them. Its generation spend cap defaults to $1.00 per run when unset, which is a ceiling and not a price.
An Agent Completion is the third path: you send the task on every call and must include generation_spend_cap_usd, which has no default. For a recap that only trims, renders and captions, a cap of $1 leaves room for one retry of every step; that sizing is a judgment call, not a documented figure.
- Own cron plus the five endpoints: $0.42 a week, no Agent turn, you choose the clips.
- Scheduled run: the Agent chooses clips; generation capped at $1.00 per run by default; LLM turn billed separately.
- Agent Completion: same Agent, task per call, cap required on every request.
Check the real spend after the first Friday
After the first run, read the ledger instead of trusting the table. GET /v1/usage?run_id=arun_... sums what that run cost, including the Agent's own turns, and the run receipt's usage.billable_amount_usd_micros shows generation spend only. If the debited figure is far above $0.42, the Agent is generating media you did not plan for, and the instructions should say to use the supplied clips only.
Sources
Related posts
More in Use cases
- A weekly host video while Tavus Griffin is preview-only: Sume, priced
Tavus describes Griffin as a preview with no date or price. For a weekly on-camera update today, Sume Avatar 1.0 costs $3.68 to $11.00 for 20 seconds.
- Wellness coach's 3-minute holiday pause video: $0.61 on Sume
A calm 3-minute pause with a slow voice, a looped music bed and one still image costs $0.61 on Sume: voice $0.13, music $0.125, image $0.05, render $0.30.
- About this ad: where Meta shows AI info for third-party AI ads
Meta's June 2026 update puts AI info in a new About this ad menu on every ad. What it says about ads made with outside AI tools, in Meta's own words.
- What is TikTok's The Next Episode, and what to have ready
TikTok's The Next Episode was reported to fund creator-led series. What was said, what was not, and a pilot kit you can build with Sume for about $157.
Written by Sume