The engine

It shows its work on your video

Most clippers hand you files and a number. This one hands you the reasoning: what each moment scored, and why the ones it skipped lost. What we do not publish is the recipe — that took a thousand hours and it stays ours.

What you get to see

The report, not the algorithm

Every finished job carries its own evidence. You never have to take our word for a cut.

01A score on every clip Each delivered short carries the score that earned it a place, so you can see which moments the engine was confident about and which were marginal. PER CLIP
02The moments we skipped Timestamped, with what was said. If you disagree with a call, you can point at it — and that is the loop that improves the engine. PER JOB
03Traceable back to your source A short is a jumpcut, so 0:25 in the clip is not 0:25 in your VOD. Each clip keeps the map back to the original timestamps. EXACT
04A report button on every clip One click when something is wrong. Reports are read, not collected — several entries in the changelog started as one. ALWAYS
The rules it will not break

What the engine refuses to do

These are commitments, not settings. They hold on every job.

It will not colour a line it is unsure of

When two people talk over each other and the match is not certain, the caption stays uncoloured. Fewer coloured lines beats putting your name on someone else's sentence.

It will not fill a quota

There is no target number of clips. A quiet recording returns few, or none, with the reasons attached. A brilliant one returns everything that cleared the bar.

It will not ship an unverified caption

The finished file is sampled and the pixels are checked. A caption that failed to render is caught before delivery, not by you afterwards.

It will not charge for a failure

A job that produces nothing costs nothing. There is exactly one place in the whole system where a customer is charged, and it runs after the files exist.

It will not put a slur in your feed

Masked in text, bleeped in audio — and moments that read as racist are removed from selection entirely, because a bleep does not save a clip that is about the joke.

It will not train on your content

Your source is deleted once you collect the results. Nothing you upload teaches the engine anything.

See it on your footage

The detector readout is free

Upload an hour and read what it measured. That is the whole argument.