What can a $5 generation spend cap make? Clips, stills, captions
A $5 cap on an Agent Completion covers 16 five-second Wan 3.0 clips at 480p, or 50 nano-banana-2 stills, but not one Seedance 2.5 clip at 720p and 15 seconds.
Agent Completions require generation_spend_cap_usd on every call; there is no default (Agent Completions). That is a feature: the caller has to decide what a request may spend. The practical question is what a number like $5 actually covers, because some prompts cannot finish under it.
The table prices common items from the Sume rate card read on 2026-10-03 and divides $5 by each price. Each row is one kind of item on its own, not a mix.
What does $5 cover?
A request can spend up to its cap and not beyond it. An item whose own price is above the cap cannot be generated within it, so the Seedance 2.5 clip at 720p and 15 seconds ($8.667) needs a larger cap.
| Item | Price each | Fits in $5.00 |
|---|---|---|
| Wan 3.0 clip, 5 s, 480p | $0.3125 | 16 |
| Wan 3.0 clip, 5 s, 720p | $0.625 | 8 |
| Wan 3.0 clip, 5 s, 1080p | $1.25 | 4 |
| Seedance 2.5 clip, 15 s, 480p, 9:16 | $4.0302 | 1 |
| Seedance 2.5 clip, 15 s, 720p, 9:16 | $8.667 | 0 (does not fit) |
| Avatar Video plus, 20 s, no product | $4.90 | 1 |
| Avatar Video max, 20 s, no product | $11.00 | 0 (does not fit) |
| Image, higgsfield-soul | $0.005 | 1000 |
| Image, gpt-image-2.5 low | $0.0075 | 666 |
| Image, nano-banana-2 | $0.10 | 50 |
| Image, nano-banana-pro | $0.1875 | 26 |
| Video caption job, up to 60 s | $0.20 | 25 |
How do you choose a cap for a mixed request?
Add the planned items and a margin. A request for six 5-second Wan 3.0 clips at 720p and six captions is 6 x $0.625 + 6 x $0.20 = $4.95, which is just under $5 and leaves almost nothing for a retry. Set the cap at $7 or $8 and the agent has room to regenerate one clip.
Remember that image prices differ by a factor of 37.5 between higgsfield-soul ($0.005) and nano-banana-pro ($0.1875). If a prompt lets the agent choose the model, a cap of $5 bounds the damage; if you pin the model, you can size the cap from the table.
Where does the cap stop?
The cap bounds generation spend for that request. It does not replace the wallet: if the balance cannot cover a job's reserve, the job returns 402 insufficient_credits. For schedules, the scheduled-run default cap is $1.00, which is a fifth of this table's budget.
Sources
Related posts
More in Agents
- What is an AI video agent? One that talks, or one that makes
An AI video agent can mean a live on-camera persona like a Tavus PAL, or an agent that makes videos for you. Which one Sume is, and how to call it from code.
- Zed ask_user forms: confirm a Sume estimate before paying
Zed's ask_user tool lets agents ask with forms. Use it as a spend checkpoint with Sume's dry_run preview, so people approve a number, not a tool name.
- Run the Sume video agent from your backend with Agent Completions
POST /v1/agent/completions runs the same agent as the Sume Agents chat, with tools and media generation, and returns an async run receipt you poll or webhook.
- Safe automation for AI agents that call paid APIs
Keep agents read-only by default, keep secrets out of logs, and on hosted MCP send an idempotency_key, preview with dry_run, and cap with max_spend_usd.
Written by Sume