Volume is the job.Consistency is the deliverable.
Nobody keeps you on retainer for one great clip. They keep you because the fortieth is as good as the first, it lands on the day you said, and it looks like it came from their brand rather than from a tool. Throughput and consistency. Two problems, and nothing solves both at once.
You are billed for the footage that goes in, and downloading what came out costs nothing however often you do it. Changing a finished clip is a re-cut - no detection, no review, no ranking pass, just the encode and the fetch - which is a few credits rather than another job. Asking for twelve clips instead of five at submit does move the number, because the review reads three candidate windows for each one.
Logo, accent colour, caption style and post voice live on the account and land as the defaults on every new project. Nobody has to remember them.
An HTTP API and an MCP server, so clipping is a step in the automation you already run instead of another tab somebody has to open.
- motion
- audio
- scene cut
- nominated
Every account looks different, which is the problem with a tool that only reads words. The same pass runs on a keynote, a walkthrough and a sizzle reel, and the review is what stops the graphics-only slate at 0:30 from being delivered to the client as a clip.
Illustrative shape · your own curves are measured, not drawn
Re-cut it again.That part is free.
One credit is 60 seconds of source video at the base rate, rounded up per job, plus a start-up charge for the work that happens once however long the file is. The expensive part is reading the footage, so that is what the bill is mostly made of. Exporting a clip you already have costs nothing, however many times you do it. Restyling, reframing or re-trimming one is a re-cut - the moment is already known, so there is no detection, no review and no ranking pass, but the video is encoded again and the original has to be fetched again to do it. That is a few credits, not a job. Charge per export and your editors start rationing their own experiments, and you can see that in the work.
Clips per job
One to twenty, ranked by score. Asking for more does cost more - the review looks at three candidate windows per clip - but the footage is only read once either way, so twenty clips is nowhere near four times the price of five. Nothing you do to a clip that already exists costs anything at all.
Re-cuts
A finished clip already knows where it is, so re-cutting is anchored work: change the trim, the crop, the caption style or the accent colour and a 45-second clip costs 18 credits.
Only the part you need
Hand it 10:00 to 30:00 of a three-hour recording and only that stretch is processed. The rest of the file is still downloaded before anything can seek into it, so it carries a small handling charge - the estimate shows it as its own line.
Failures
A failed or cancelled job returns the hold in full. The usage ledger shows every movement against the job that caused it.
The meter is on the input, which cuts both ways. A three-hour file handed over whole costs 477 credits even if you keep ninety seconds of it. The same file with a range set to 10:00-30:00 costs 128. Pasted timestamps do not shrink it the same way - anchors take precedence over a range, so an anchored three-hour job still bills all three hours, at 178. Set the range.
A month ofclient work.
Four clients, four kinds of job, the volumes a small shop actually runs. The volumes are illustrative. The credit figures are not - they are computed on this page by the same function that takes the hold on a real job.
| Work | Source | Each | Runs | Credits |
|---|---|---|---|---|
| Weekly interview show45 minutes of source, full review, cut to six verticals | 45 min | 187 | 4 | 748 |
| Monthly event recapTwo hours of stage and floor footage, one pass | 120 min | 348 | 1 | 348 |
| Social shoot daysShort vertical-first sessions, twice a week | 15 min | 119 | 8 | 952 |
| Webinar with a run of showTimestamps pasted in, so detection is skipped and it bills at the anchored rate | 60 min | 60 | 3 | 180 |
| 16 jobs, every clip they produce | 2228 |
That is 2,228 credits against the 2,400 in Studio, $249 a month. The 172 left over buy about 38 minutes more source, or 9 re-cuts. The re-cuts are what the headroom is really for - the first automatic cut of a client deliverable is rarely the one you send.
Read that table for the surprise in it: social shoot days is the biggest line, at 952 credits, even though it is the shortest footage on the list. Half of what a job costs happens once - the review looks at the same shortlist whether the source is fifteen minutes or fifty - so eight short uploads pay that eight times. Batch a shoot day into one file and the same footage costs a fraction of it.
The API and signed webhooks are on Studio, $249 a month with 2,400 credits. If you are driving Cutlist from your own scheduler rather than the studio, that line is what you are buying. The credits are the smaller half of the decision.
One brand kit,per account.
Consistency across a client roster is the thing an agency actually needs, and this is where the product is short. The kit - logo, accent colour, default caption style, default post voice - belongs to the account. It does not belong to a client, and there is no shelf of six to switch between. Here is exactly what that costs you, and what people do instead.
In the studio
Caption style and post voice are picked per upload, so those two travel with the client without touching the kit. Accent colour and logo come from the account and land on every render.
In the editor
A finished clip can be re-cut with a different accent colour and caption style at the anchored rate. It is a repair, not a system, but it is cheap enough to use as one.
Through the API
Every brand field is a request field: captionStyle, accent, logoUrl, logoPosition, logoOpacity, logoScale. A shop driving jobs programmatically has real per-client branding today, and the account kit is only the default it overrides.
Separate accounts
One account per client works. It also splits your credits into pots that cannot be pooled and gives you several bills, which is usually a worse trade than swapping a logo.
There are no team workspaces and no per-seat roles. One account, one login, one set of projects - so a shop with four editors is sharing a credential. That is the real cost of this gap, not the logo.
Clipping as a step,not a destination.
Submit a job over HTTP and take a signed callback when the clips are ready, or point an agent runtime at the MCP server and let it call cut_video the way it calls anything else. Same engine, same credits, same output - nobody sits and watches a progress bar.
Submit and forget
POST a source URL and options, get a job back. Poll it, or let the webhook find you when the render is done.
Your ids come back untouched
Send a client id or a job number in metadata and it returns alongside the clips, so your handler keeps no state of its own.
Four MCP tools
cut_video, get_clips, estimate_cost, list_jobs. estimate_cost exists so an agent can ask the price before it spends anything.
Per-job branding
Caption style, accent colour and logo placement are request fields. That is how one integration serves a whole roster.
30 days,then it is gone.
Clips live 30 days and then delete themselves. Source uploads go within 2 days of a finished render. Deliberate, not an apology: a clip you have not delivered in a month is not a deliverable, and we would rather not quietly become your archive.
100 GB per account
Shared across every project, with usage shown in the studio. An allowance, not a library.
Everything in one zip
Every clip and every caption file for a project, named from the clip titles. Deliver straight out of it.
A countdown on every clip
Each one shows the days it has left, so nothing disappears on you unannounced.
There is no client review layer - no approval queue, no comments, no per-client portal. You download and you deliver, the way you already do.