This finds them.
Upload an hour of talking and Heyper Clips reads every word, picks out the moments that stand up on their own, and scores each one four ways so you can disagree with it. Every clip starts and ends on a whole word, sits in a vertical frame, and arrives captioned.
- scores per clip
- 4scores per clip
- vertical by default
- 9:16vertical by default
- every start and end
- Whole wordsevery start and end
- before you download
- Captionedbefore you download
Free to start · no card required
Shown their working.
Click a clip on the timeline, or re-rank by whatever matters most to you today. Every score you see is worked out exactly the way the app does it — including the sum underneath.
“We fired our best-performing channel.”
8.7
We fired our best-performing channel. On paper it was half our signups. In practice it was half our signups and none of our revenue, and it took us fourteen months to notice the difference.
Does the first line stop someone scrolling?
Is something real being said?
Does it make sense on its own?
Would anyone send it to a friend?
A number that contradicts itself, then the resolution. Self-contained, and the kind of thing people quote.
9.6×0.4 + 7.4×0.25 + 8.2×0.2 + 9.1×0.15 = 8.7
gives you
Moments, not chunks
It looks for moments that make sense on their own — a point made and landed — instead of chopping the hour into equal pieces and hoping.
Four scores, not one number
Hook, flow, idea and shareability, each rated on its own. “Brilliant opening, unlikely to be shared” is something you can act on; a single 7 out of 10 is not.
We check the AI's homework
It suggests the moments. We then line both ends up with whole words, check the length, remove overlaps and work out the score ourselves. AI picks what is interesting; it does not get to do the maths.
Vertical by default
Ready for Reels, Shorts and TikTok, with square, portrait and landscape there when you need them.
Captioned on the way out
The same captions as Heyper Video, timed to each word — so a clip is ready to post the moment it finishes.
Your call on length
Under thirty seconds, thirty to sixty, sixty to ninety — or leave it open and let the material decide.
one honest total
Shown separately because they often disagree — and the disagreement is the useful part.
- Hook40%
- Does the first line stop someone scrolling?
- Idea25%
- Is something real being said?
- Flow20%
- Does it make sense on its own?
- Shareability15%
- Would anyone send it to a friend?
The AI gives its opinion on each of the four. It is never asked for the final number — we work that out ourselves, so the total always adds up to the four scores you can see.
Not with the clock.
AI is good at spotting the moment that lands, and surprisingly bad at getting the timecode right. So we ask it for the first and check the second ourselves.
Every suggestion gets verified
The AI says where it thinks a clip starts and ends. We then move both ends onto whole words, check it is a sensible length, remove any overlap with another clip, and work out the final score ourselves.
Nothing is made until you say so
A clip starts life as a suggestion, not a video file. Running the search again on an hour of footage costs you a little thinking time and nothing else, and a clip you scroll past costs nothing at all.
One number would hide the argument
“Great opening, unlikely to be shared” is something you can act on. A single 7 out of 10 tells you nothing about why it ranked where it did, or whether you should ignore it.
Ask for what you actually need
Add a note — “the part about pricing”, “anything with a number in it” — and the search leans that way. Leave it blank and you still get results.
Does the work three times over
The same timings that decide where a clip starts also caption it and strip the “ums” and dead air out of the middle of it.
The editor is there for the awkward one
Any clip opens in Heyper Video with its words intact — for the one that needs a different crop, not as a compulsory last step for all ten.
applied to every clip
You choose how clips should look before the search runs. Ask for ten captioned vertical clips and that is what lands — not ten half-finished ones waiting on you.
Under 30s · 30–60s · 60–90s · Any length
The lengths people actually post, rather than a box to type a number into.
9:16 · 1:1 · 4:5 · 16:9
Vertical by default, framed around the person speaking rather than the middle of the shot.
Fillers · pauses · censor
All three come free with the transcript — no extra processing and no extra waiting.
12 styles
The same captions as Heyper Video, timed to each word and baked in.
hours of good material
The clips were always in there. Finding them was never the creative part of the job.
An episode becomes a week of posts
Publish the long cut, then let the ranking decide which four minutes of it go out on Monday, Wednesday and Friday.
Salvage the ninety-minute recording
Nobody watches the replay. Six vertical clips from it, captioned and ready, get watched.
Find the moment the customer said it
Point it at a recorded call and get back the bit worth sending to the team, cut cleanly at both ends.
Stop scrubbing your own footage
You already know the good bit is in there somewhere. This is the part of the job that was never creative.
A lecture, indexed by idea
The windows that stand alone are exactly the ones worth pulling out as a revision clip.
Volume without a second editor
One long video per client per week turns into a month of posts, without anyone opening an editor.
Upload a long video and see what it finds.
plainly
- Scoring
- Hook, flow, idea and shareability — weighted into one total
- Clip length
- Under 30s, 30–60s, 60–90s, or any
- Shapes
- Vertical, square, portrait and landscape
- Start and end
- Always on a whole word, never mid-syllable
answered
It reads the whole transcript and suggests moments, with a reason for each one. Everything it suggests is then checked: the start and end move onto whole words, the length is checked, any overlaps are removed, and the final score is worked out from the four parts rather than taken on trust.
Hook is whether the first line stops someone scrolling — it counts for the most, because nothing else gets seen without it. Flow is whether the clip makes sense on its own. Idea is whether something real is being said. Shareability is whether anyone would send it to a friend.
No. Both ends always land on a whole word, so a clip never opens or closes on half a syllable.
Yes — a clip is a suggestion, not a finished file, so nudging one costs nothing and running the search again does not throw away any work.
Yes, with the same word-by-word captions Heyper Video uses — so what you get back is ready to post.
with
Your next frame is one prompt away
Open the prompt box, pick a model, and see what comes back. No card, no setup, no onboarding call.