A five-step edit chain on Sume costs 27 cents: trim to render

Run one clip through trim, filter, audio concat, compose and a two-minute render on Sume and the whole chain costs $0.27: $0.02 + $0.02 + $0.01 + $0.02 + $0.20. None of these steps uses provider inference. They are worker ffmpeg jobs at fixed per-job rates, except the render, which is $0.10 per output minute rounded up.

A thousand such chains is $270, and the render dominates at 74 percent of the total. If the finished video is 60 seconds or less, the render drops to $0.10 and the chain to $0.17.

The line items

Rates are the public Sume prices, read 2026-10-09.

Free checks before paid steps

Two of these steps have a free dry run. POST /v1/video-filter/check runs the same checks as the encode, returns diagnostics and an estimate, and does not create a job or reserve credits. POST /v1/timeline-1.0/plan returns billable_minutes and estimated_cost_usd_micros without a job. A program that passes a check can still fail on the worker, for example on a bad expression, memory or time, and then you get a structured job error.

Use the checks in a pipeline to stop a bad program before any of the five steps are paid for. At $0.27 a chain the saving per caught error is small; across a batch of thousands with a templating bug, it is the whole batch.

Where the chain can cost more

The per-job price of a step does not change, but the number of jobs does. Each step is a separate job, so a chain with three filter passes is $0.31 rather than $0.27, and a 121-second render moves from two minutes to three, adding $0.10. Media must also be this workspace's media.sume.com artifact or asset, so a source URL elsewhere needs a media import first; the import is its own step, and the docs here do not price it, so check the live catalog.

Because these are many short jobs, capacity is about job count rather than money: a 1,000-chain batch is 5,000 jobs, and Pro accepts 24 at a time. Pace the submits.

Chain cost at scale

Multiply the chain by volume: 100 videos a day for 30 days is 3,000 chains, $810 at $0.27, of which $600 is renders. Shaving one minute from each render by cutting the output from 120 to 60 seconds takes the month to $510. Because every step is a fixed per-job rate, the budget is a plain function of job count and output minutes. The risk is not price but retries: a template bug that fails the last step after four paid steps wastes $0.07 per chain, and the free checks exist to catch that earlier.