Benchmarks
How Clypmint does on the jobs people ask for, next to other AI video editors. A number goes on this page only after we measure it, with the method, the date and a link to the data. Until then the row says Pending.
Updated October 20, 2026. Nothing is measured yet, so every row says Pending. Read this page as plain text.
How we measure
Every row uses the same footage and the same request for every tool. The request is written before the first run, so nobody can tune it to win.
- We use our own footage for these tests, never a customer's video.
- The request is written once and sent to every tool word for word.
- Times start when the upload finishes and stop when the first edit can play.
- We report the median and how many runs it came from.
- In side-by-sides, people pick between two edits without knowing which tool made them. The order is shuffled.
- If a tool cannot do the job, the row says so. We do not leave it blank.
- A row with a number has a link to its data. A row without a link is not measured.
- We make Clypmint, so we are not neutral. That is why the method is on this page.
Compared with: ChatCut, Vyra, Captions.
The rows
| What we measure | Clypmint | ChatCut | Vyra | Captions | Source |
|---|---|---|---|---|---|
| Time to a first playable editLower is better. A talking video of up to 3 minutes. The clock starts when the upload finishes and stops when the first edit plays. | Pending | Pending | Pending | Pending | Replay of real requests on staging, one run record per run |
| Time to apply a simple change you typeLower is better. Requests such as make the opening faster or smaller captions, sent on a finished edit. The clock stops when the changed version can play. | Pending | Pending | Pending | Pending | Replay of real follow-up requests on staging |
| Share of your asks that the edit keptHigher is better. We split each request into its asks: format, length, text on screen, audio rules and your own media. Then we check each ask in the finished video. | Pending | Pending | Pending | Pending | A fixed set of real requests, checked by code and by eye |
| Requests refused instead of repairedLower is better. A request counts as refused when the tool stops with no video and no usable answer. | Pending | Pending | Pending | Pending | The same set of real requests, run on staging |
| How often people pick the Clypmint edit in a blind side-by-sideHigher is better. Two edits of the same footage from the same request, side by side, with the tools hidden. Ties count half. We publish the number of votes with the result. | Pending | Pending | Pending | Pending | Blind votes on our own test footage |
| Price of a one-minute editLower is better. What one run costs you at list price, on the plan it needs. For other tools we use the prices they publish and say when we read them. | Pending | Pending | Pending | Pending | Plan prices and the credits each run used |
| Time to cut clips from a long videoLower is better. A 60-minute recording, asking for the best clips that stand alone. The clock stops when every clip can play. We also report how many clips people judge ready to post. | Pending | Pending | Pending | Pending | Test recordings that we own |
| Cuts that land on the beatHigher is better. A montage from many clips to one song. A cut counts as on the beat when it lands within one frame of a beat we marked in the song. | Pending | Pending | Pending | Pending | Montage test set with marked beats |
| How close an edit is to the reference it copiesHigher is better. We measure cut lengths, caption position, colour and motion in the reference and in the result, then report the gap. People also vote on which edit feels closer. | Pending | Pending | Pending | Pending | Measured reference set |
| Time from the install prompt to a first edit in ChatGPT or ClaudeLower is better. We start with no plugin installed. We paste the install prompt and stop the clock when the first edit plays. | Pending | Pending | Pending | Pending | Timed runs in ChatGPT and in Claude |
Pending means we have not measured it. We do not fill a pending row with a guess.
See the work
Numbers are one half of proof. The other half is the video. See real edits in the showcase, or try Clypmint yourself. Free to make and watch, a paid plan to download.
Judge it on your own video.
Free to make and watch, a paid plan to download.