Use case

An AI clip generator that only keeps whole stories.

MakeRoll’s clip finder reads the whole transcript of your video, proposes the moments worth posting, checks that each cut tells one complete story, and drops the ones that do not. What reaches you is a list of whole clips, each scored for how strong it is, with the quote it rests on and the reason it was picked one click away.

The job, in plain words

You want the best moments of a video without watching it again, and you want to trust the list enough to post from it. The usual disappointment with AI clipping is not that it misses a moment; it is that it hands you a clip that starts in the middle of a thought, or stops a sentence before the punchline, and you only notice after posting.

So the job is two jobs. First find the strong moments, anywhere in the video. Then make sure each cut is a moment someone can understand with no context: the setup is in, the payoff is in, and it is one story rather than two glued together.

How it works

It reads everything
Your video is transcribed word by word, with each word timed to the moment it is spoken, and the finder reads the whole transcript, the last hour as closely as the first. It looks for moments that work on their own: a story, an answer, a reveal, a strong opinion, and moments where the picture carries the point.
It checks every cut is a whole story
Each proposed moment is cut to sentence boundaries, then checked by a separate pass: does it have its setup, does it reach its payoff, is it one story? A cut that fails is re-cut once and checked again. If it still is not whole, it is dropped rather than shown to you with a low score.
You see the score and the reason
The clips that pass arrive with a score between 60 and 96 for how strong the moment is, a title and a suggested hook headline. Open a clip to read the quote it is built on and why it should hold attention, written in the clip’s own language. Then like, skip, edit or schedule.

Example: what “whole” means on one interview

Imagine a 40-minute interview with a chef. Three cuts show how the check works:

Kept as proposed: the chef is asked about her worst night in a kitchen, tells the story, and ends on what she changed the next day. Question, story, ending: whole.
Re-cut: the first proposal began on “and that’s why I never trust a supplier again”. Without the story before it, the line means nothing, so the cut is widened back to where the story starts.
Dropped: a passage that jumps from pricing to her childhood to a recipe. Every piece is interesting, but no single cut makes one story of it, so it does not reach your list.

This example is illustrative, to show the rule rather than a recorded run. The rule itself is real: MakeRoll delivers only clips judged complete, which is why a video can give fewer clips than you expect.

How to read what the finder gives you

The score is strength, not completeness

Every delivered clip already passed the whole-story check. The score, from 60 to 96, says how strong the moment is compared with the others, not whether it is finished.

Fewer clips is a feature

A slow video gives fewer clips because there are fewer complete moments in it. Padding the list with broken cuts would only move the sorting work back to you.

Read the reason before you post

The analysis names the concrete moment the clip relies on. If that moment is not what you would sell the clip on, re-cut it or skip it.

You still have the last word

The finder proposes; the editor decides. Move a start or end by editing the transcript, or add a moment it skipped in Cut mode, at no cost.

What you end up with

Only whole stories

No clip that starts mid-thought, stops before the payoff or splices two stories together reaches your list.

A score you can sort by

Scores run from 60 to 96, so sorting by score puts the strongest complete moments first.

The reasoning, on click

Each clip shows the quote it is built on and why it was picked, so you can judge it in seconds.

What this does not do

The finder works from what is said, plus the picture where the picture is the point; a video with little speech gives it less to go on. It suggests titles and hook headlines, but it does not predict views, and a score is not a promise of reach.

Questions people ask

How does MakeRoll choose which moments to clip?

It reads the full, word-timed transcript of your video and proposes moments that stand on their own. Each cut is then checked for a setup, a payoff and a single story; cuts that fail are re-cut once and dropped if they are still incomplete.

What does the score mean?

How strong the moment is as a standalone clip, on a scale the finder keeps between 60 and 96. It ranks complete clips against each other; it is not a forecast of views.

Why did I get fewer clips than I expected?

Because MakeRoll only delivers clips judged complete. A video with fewer self-contained moments gives fewer clips, and you can add any moment you want yourself in Cut mode.

Does it work in languages other than English?

Yes. Transcription works in French, English, Spanish, German and dozens of other languages, and each clip’s analysis is written in the language spoken in that clip.

Does the AI clipping cost extra?

No. It is included in the credit charged for each minute of source video, on every plan, the free one included.

Credits are source minutes: a 60-minute video costs 60 credits, however many clips come out of it. Editing and exporting are free and unlimited. Pricing

Try it on your own footage.

An hour of source video free every month: full pipeline, full editor, no watermark, no card.