Picture Lock / 04 · Styles
608 styles, measured off real video.
26 things measured across 6,677 real videos — cut rate, framing, grade, palette, how long a shot is held — rather than a designer's idea of a look.
A style here is a set of numbers taken off real footage, not a preset somebody drew. 26 are measured in all; 18 of them sit on every one of the 608 styles, and the rest only where the footage allows it — a silent clip has no sound texture to measure, and saying so is cheaper than filling the gap with a number.
What you can make today
The 608 are the library the engine measures against and scores your render on. These are what you actually pick in the app — each one renders today, and the description is the one the picker itself shows you, not a nicer version written for this page.
| Style | What it does |
|---|---|
| Fully edited | Hook in the first second, dead air cut, purposeful punch-ins, word-by-word captions with a keyword, b-roll where it clearly matches, a light bed under the voice. |
| Talking head, minimal | Dead space cut, 2–3 word captions, subtle bed, a text hook. No zooms, no sound effects. |
| Raw and organic | Real footage only, natural colour, tight cuts, gentle punch-ins at real cuts, plain captions. No music, no graphics. |
| Pop-up graphics | Upload yourself talking. A hook title in the first two seconds, graphics that pop up on the words you actually said, numbered points when you count, punch-ins and subtle sound. Captions picked for the video. |
| Words behind you | Huge bold words behind you as you talk: you’re cut out frame by frame and stand in front of them. Made for podcasts and slower talks. |
| Desk | You, cut out on a dark grid. When a point lands you step to the bottom corner and the graphic takes the top: alert banners, numbered cards, the number you said. |
| Documentary trailer | Cold open on the tension line, two or three escalating points over b-roll, the music drops on the reveal, an unresolved end card. |
| News explainer | A link or an article becomes a 30–45 second explainer: a script written only from the story, a voice, and animated cards — no footage. Every line shows the sentence it came from. |
Four more run on their own tool rather than from the picker, so they are listed as what they are rather than counted as something you can press a button for.
| Style | What it does |
|---|---|
| Scroll film | A website becomes one locked camera: the page builds itself on screen while a voice reads what’s actually there. |
| Kinetic type explainer | Off-white canvas, words arrive one at a time in two weights, one tilted screenshot on a soft shadow. Every shot cuts, nothing fades. |
| ASMR product | Slow push-ins on texture, true colour, no music, no captions — the sound of the thing is the whole video. |
| Vox pop | Quick answers from different people to one question, hard cuts between faces, the question on screen throughout. |
What gets measured
- how often it cuts
on all 608 - how long a shot is held
on all 608 - hard cuts or blends
on 590 - whether it is one take
on all 608 - cuts landing on the beat
on 410 - the jump-cut feeling
on all 608
- handheld or locked off
on all 608 - punch-ins
on all 608 - how close the framing sits
on 524 - where the eyes land
on 524 - shot scale
on 398 - split screen
on all 608
- contrast
on all 608 - saturation
on all 608 - warmth
on all 608 - how crushed the blacks are
on all 608
- whether captions are burnt in
on all 608 - how much type is on screen
on all 608 - where the captions sit
on 386
- loudness
on all 608 - the sound texture
on all 608 - how much of it is speech
on 38
- whether anyone is on screen
on all 608
A style, as it is actually stored
Three real entries from the library. Nothing here is a label someone chose — every figure was read off the footage.
| cuts/min | shot held | shake | captioned | no presenter | videos | |
|---|---|---|---|---|---|---|
| The fastest in the library | 215.1 | 0.35s | 0.110 | 14% | 97% | 14 |
| The median style | 15.9 | 4.55s | 0.004 | 93% | 2% | 14 |
| The slowest in the library | 0.0 | 11.29s | 0.000 | 57% | 90% | 14 |
Five of them are measured and never scored
where the captions sit, how long before the first cut, loudness, whether it is one take, shot scale are read off every style and reported, but they are not scored against your render, because the engine has no honest lever for them. Measuring something is not the same as being able to reproduce it, and the difference is printed rather than hidden.
Every figure on this page is read
straight out of the style library by build_styles_page.py — last read 27 September 2026.