Rewound
Clips cut where viewers went back

Paste a YouTube link. Three vertical clips come back, cut at the moments viewers replayed most.

No account, no card. Three clips, free.

Pricing Counted in clips you keep · cancel any time
Freeno card
$0
3 clips, once
Full quality, no watermark. Enough to find out whether the graph knows something you didn’t.
Creatorstart here
$10/mo
100 clips · 10¢ a clip
One channel, posting steadily. Upload your own footage as well as pasting links.
Studiobest rate
$29/mo
400 clips · 7.3¢ a clip
A back catalogue, or several channels at once. Priority in the queue.

Nothing to install · no password · your plan is attached to the email you pay with · billed monthly in USD by Stripe · cancel from the first email

The moment-finding problem

Everyone else guesses which thirty seconds are good.

Paste a link. Three clips come back. Free, no account, no card.

YouTube already knows which parts are good. Every video carries a replay graph built from people dragging the scrubber backwards. Rewound reads that graph and cuts where the peaks are.

Example output 9:16 · captions burned in

Real output, cut by Rewound from full-length episodes. Each clip credits its source creator on screen.

Replay intensity · one video 30 Years of Business Knowledge  2:26:13
14:37Highest replay density anywhere in the video
1:04:20Second peak, fifty minutes clear of the first
1:55:31Third, trimmed back to a sentence boundary
How it works

Three steps, none of them editing.

The replay data arrives with the video's metadata, before a single frame is downloaded. So it knows where to cut before it starts.

Step one

Read the graph

Smooth it, then throw away the opening spike. Every video's first seconds are its most replayed and they are never the interesting part.

Step two

Cut on the peaks

Highest non-overlapping moments, weighted backwards toward the setup that precedes each payoff. In and out points snap to sentence boundaries.

Step three

Ship it vertical

Reframed to 9:16, captions burned in word by word, source credited on screen. A title, description and hashtags come with each one.

What the others do

A language model reads the transcript and estimates which passage sounds compelling. It is a guess about text, made without ever seeing how one viewer behaved.

What Rewound does

Reads the record of what viewers actually did. Not a prediction of what should have been interesting. Evidence of what was.