Sume flat job prices: trim, filter, compose, detach, transcript

Most of Sume's post-production jobs are priced in single cents, because the trim, filter and timeline docs say they run on worker ffmpeg with no provider inference. As of 2026-10-04 the docs list $0.01 or $0.02 for the trim, filter, compose, audio and detach jobs, with captions at $0.20 as the dearest flat item.

The price list

Each rate is as printed in the model docs. The docs tell you to confirm live prices in GET /v1/catalog.

A full clean-up pass

Trim a clip ($0.02), filter it ($0.02), detach its audio ($0.01) and render a one-minute timeline ($0.10) for $0.15 of platform jobs. Add captions and it is $0.35. That is well below the cost of regenerating the footage, so edit before you regenerate.

Why cheap jobs matter to a budget

Regenerating a 5-second shot at Wan 720p costs $0.625. Trimming a good clip to the best 3 seconds costs $0.02. If an existing clip is nearly right, a trim, a filter or a recompose is 30 times cheaper than a new take. The platform rates are not why a video costs what it does, but they are why editing is the first thing to try.

Each job is reserved and captured as its own ledger row, so a pipeline of five small jobs produces five rows in the usage list. Read them by run or by job id when you want a per-video total.

Notes

• The video filter docs say the validation check is free.
• Without a duration_seconds hint, a transcript reserves 1 minute.
• A video inspect that transcribes also reserves its compute ceiling, so its capture is the container seconds at the Modal list price times 1.25 plus the platform fee, never above the hold.