How much does it cost to try an AI video API?

Trying an AI video API costs between $0.12 and $1.08 per clip on Sume for the five models in the table, sent at each model's shortest length and lowest resolution. Those are the amounts Sume reserves, before a 5.5% agent fee by default.

Sume bills routed video at the provider's list price × 1.25 (Video generation). The amounts are computed from Sume's pricing code and rounded up to the cent, as read on 2026-09-29.

What is the smallest test clip?

A test needs to prove your plumbing (submit, poll, download), not the final look, so use a minimum length and the lowest resolution. The lengths and resolutions below are the minimums the docs list for these five models; the other listed video models are not priced here.

Why do the test clips differ so much in price?

Three things move the amount: the model's rate, the resolution and the length. wan-3.0, gemini-omni-flash-1.1 and the MiniMax models charge per output second at a rate set by resolution, so a shorter clip at a lower resolution costs less. The Seedance models charge per video token, and the token count grows with width, height and seconds, so a small frame and a short length keep the amount down.

Send resolution on every test request. For wan-3.0, leaving it out reserves at the 1080p rate, the highest of its three.

How much do I need in the wallet first?

Enough to cover the reserve of the clip you submit. Sume's pricing FAQ says teams can fund the wallet with on-demand purchases from $10 to $1000, so the smallest purchase covers many test clips.

Top-ups happen in the dashboard's Billing & subscription page; the public API only reads balance and usage. If the balance cannot cover the reserve, the submit fails with 402 insufficient_credits before any provider work starts. See Billing and credits.

What does a first request look like?

Send model, prompt, duration and resolution to POST /v1/videos, with an Idempotency-Key so a retry cannot create a second job. Poll the job, then download the file.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: first-test-001" \
  -d '{
    "model": "wan-3.0",
    "prompt": "A paper boat drifts across a puddle, soft daylight.",
    "duration": 2,
    "resolution": "480p"
  }'

Does a failed test cost anything?

Not as spend. Sume reserves the amount when the request is accepted, captures it when the job completes successfully, and releases or refunds the reservation when a job fails. A queued job you cancel before generation starts is also released; once generation has started, the cancel call answers 409 job_generation_already_started. The details are in Generation admission and Do failed AI video jobs cost credits?.

How do I check what a test actually cost?

Call GET /v1/usage with job_id. The response adds a summary, and debited_usd is the figure to quote; held and refunded amounts are not spend. See Usage. Only the clip you keep at your real resolution and length tells you the production cost, so price that setting with How much does an AI video cost?.