Clipping software that cuts nothing you have not approved
It reads the whole video, ranks every passage worth cutting and shows the sentences that made each one a candidate. You keep or you pass — a plan is metered in credits, spent on the minutes it reads and, past a generous included batch, the clips you export.
What one run is made of
How it choosesVideo clipping is two jobs, and only one of them grows with the video
Reading an hour of footage costs about twice what reading half an hour costs, every time. Cutting a forty-second clip out of it costs the same whether the episode ran ten minutes or four hours, because the renderer only ever fetches the seconds you approved.
No card
30 credits, granted once — about 30 minutes of video, 10 rendered clips included
$19 a month
240 credits a month — about 4 hours, any length. 80 rendered clips included, then 3 credits each
$49 a month
660 credits a month — about 11 hours, any length. 220 rendered clips included, then 3 credits each
$99 a month
1,620 credits a month — about 27 hours, any length. 540 rendered clips included, then 3 credits each
A clip costs the same, whatever the video
A tool sold by the export has to guess how many you will take, then price for the worst case. This one ties the free batch to how much video you read — a generous twenty clips an hour — and only past it does an export cost anything: a flat three credits, never scaled to the length of the source.
Reserved, then corrected
The allowance is charged against the probed duration when the job is queued, then reconciled to the measured one after transcription — a container’s declared duration and its real one disagree often enough to matter.
A failure is not a purchase
An analysis that fails gives its minutes back once the retries are exhausted, and the ledger keeps the line. Renders never cost minutes in the first place.
Every plan runs the same selection pass, the same four pillars, all five shapes and both caption styles. What a larger one buys is hours, how many jobs may run at once, how long the library keeps what you made, and how many videos a day may arrive by link — three on the trial, twenty on Starter, fifty on Creator, a hundred on Studio. Uploads are not capped at all: pasting a link costs you nothing and costs us a fetch.
What it does before you have decided anything
A two-hour transcript is split into windows of about thirty minutes, read independently and in parallel with three minutes of bleed on each inner edge, and then a single judge ranks the pooled finalists. One pass over the whole thing anchors on what it read first and never picks from the back half — the windows are the fix for that, and the intro is detected and left out before any of it is ranked.
Two ways in, and what is read off each
An upload is probed for its duration in the browser before a byte is sent, and gets a frame grabbed a few seconds in for the poster. A link — youtube.com or youtu.be, and no other site — has its duration read from the video’s metadata instead, before anything is queued or charged, and the import does not start until you confirm the video is yours to use.
Audio first, video last
Transcription reads an audio-only stream. No copy of a link is kept here: the only video bytes ever read are the seconds inside a clip you approved, fetched per render — about six megabytes — and cached, so exporting the same moment again fetches nothing.
Candidates, not exports
Each moment arrives with four scores, its timecodes and the sentences it is made of. Keep it, pass on it, or drag the in and out points yourself.
Then it cuts
Speaker tracking follows whoever is actually talking, the banner guard punches in only if it measures a lower third baked into the source, and an MP4 lands in your library.
Two looks, both opinionated on purpose
Most clipping software answers this with a settings panel. This one ships two looks by default, because both were built to be looked at rather than configured. Pick one — per render, or from a saved preset — and every moment you approve comes out the same way.
- Boxed 1:1, with the composited text sitting behind the speaker — the subject is matted out and occludes the words
- Full-frame 9:16 with karaoke captions, highlighted word by word off the transcript
- The frame follows whoever is actually speaking, not whoever happens to be on camera
- The banner guard measures the source and tightens the shot only when it finds a chyron cooked into it
Where it stops
Three things this software does not do, listed here rather than discovered in week two.
There is no TikTok, Instagram, Twitch or Facebook integration, and nothing is posted on your behalf. Reading a video from YouTube and posting one back to it are two different connections, and only the first exists here. The finished MP4 downloads in one click and the upload stays yours.
The compositing geometry is written against the 1:1 box. The options panel corrects that pairing when you pick it, rather than letting a render fail twenty minutes in.
No tracks, no keyframes, no transitions. The decisions this software gives you are which moments to keep, where they open and close, and which shape they ship in.
Questions
Written against what the engine does, including the parts that would be easier to leave out.
What exactly am I paying for?
Credits, spent on two things. Transcription and the selection pass both cost in proportion to how long a video is — a credit a minute — and rendering does not, because the renderer only ever touches the seconds you approved. A generous batch of clips is included with every plan; only past it does an export spend three credits from the same balance.
Is there a clipping app to install?
No. It runs in a browser, and the transcription, the selection pass and the render all happen on our machines — so a four-hour episode does not tie up your laptop, and there is nothing to keep updated.
What does the free trial actually include?
Thirty credits, once, without a card — about thirty minutes of video. Ten clips are included, and so is every render style. It does not renew, because a free allowance that came back every month would be a bill that came back every month.
Does it download the whole video?
A link is fetched in full exactly once, when the first analysis runs, and never again; an upload is already here, so nothing is fetched at all. After that, transcription reads an audio-only stream off the stored copy and deletes it as soon as the words come back, and the video itself is read one clip window at a time, once you have approved that window.
What happens if a job fails?
An analysis that fails after its retries returns the minutes it reserved, and the movement is written to the usage ledger line by line. Renders are never metered at all, so a render that fails costs nothing by construction.
Can it work through a back catalogue in one go?
No. It reads one video at a time per job, though a plan can run more than one job at once — two on Starter, three on Creator, five on Studio. There is no cap on how long a single video may be any more; a four-hour episode simply spends four hours’ worth of credits.
The same engine, from wherever your video is now
Each of these pages describes one route in and says what it leaves to you.
YouTube clip maker
Paste the watch URL and it is fetched once on our side, or upload the file. Only the passages you approve are ever cut.
Podcast clip maker
Two hours read in thirty-minute windows with three minutes of overlap, so the back half is ranked too.
TikTok clip maker
Vertical 9:16 with word-by-word captions. It hands you the MP4 — the upload to TikTok stays yours.
Twitch clip maker
Export the VOD, upload it, and read a whole stream for the passages worth cutting. No Twitch account is connected.