wizz-method 2.16.0 → 2.17.0

package/src/skills-lib/wizz-video-styles/references/audio.md

@@ -0,0 +1,54 @@
+# Music and sound effects per style
+
+The measurements behind this page: [measured-patterns.md](measured-patterns.md). Per-event SFX mapping, ducking and the loudness gate belong to `wizz-video-effects`; this page decides the **bed and the budget** the style asks for.
+
+## Per style
+
+| Style | Music bed | Bed in the kit | SFX budget |
+|---|---|---|---|
+| motion-editorial | continuous, calm, ~0.2 of the voice | `piano-editorial`, `lofi-chill`, `classical-light` | soft whoosh on the 2 or 3 biggest transformations; light typing under one typed pill |
+| explicador-animado | continuous, scored | `cinematic-tense`, `tech-pulse`, `corporate-light` | whoosh on some transitions, soft UI sound on highlights |
+| meme-rapido | continuous, upbeat | `meme-beat`, `tech-pulse`, `lofi-chill` | whoosh/pop on a minority of cuts; `record-scratch` once as a punchline |
+| talking-head-legenda | none or faint | `lofi-chill` very low, or nothing | an occasional pop on a caption keyword |
+| talking-head-cards | faint | `corporate-light`, `lofi-chill` | `swoosh-text` when a card enters, not on every card |
+| podcast-clip | none or faint | nothing | none |
+| camera-no-setup | none or continuous | `tech-pulse` low, or nothing | `click-mouse`, `typing-short` only when the action is on screen |
+| quadro-branco | none or faint | `piano-editorial` very low, or nothing | none; the pen is the sound |
+| desafio-numero | faint | `acoustic-lifestyle`, `meme-beat` | `pop-bubble` or `ding-notify` when a tag or stage icon appears; `ka-ching` on the final price |
+| macro-asmr | mixed: beat (beauty) or none (recipes) | `lofi-chill` for beauty; nothing for recipes | the process sound (sizzle, pour, click) instead of added SFX |
+| pov-vlog | continuous or faint | `acoustic-lifestyle`, `lofi-chill` | none |
+| tour | none | nothing | none |
+| experimento | faint or continuous | `cinematic-tense` low | `riser-short` before the reveal, `boom-impact` on it, once |
+| esquete | none or faint | nothing | none; timing is the joke |
+
+"Continuous" means the bed runs under the whole piece with a fade-in and a 2 s fade-out; "faint" means audible only in pauses.
+
+## The kit (`assets/audio/`)
+
+Redistributable with the project: CC0, public domain, or CC BY 4.0 with attribution. Every file's source, licence page and edits are in [`assets/audio/CREDITS.md`](../assets/audio/CREDITS.md); durations and loudness per file in `assets/audio/kit.json`.
+
+- **Music** (70 s each, fade-in 0.5 s, fade-out 2 s, -20 LUFS): `piano-editorial`, `lofi-chill`, `tech-pulse`, `corporate-light`, `cinematic-tense`, `acoustic-lifestyle`, `classical-light`, `meme-beat`.
+- **SFX** (0.6 to 2.2 s, -16 LUFS target, some lower because of the -1.5 dBFS peak limit): `whoosh-soft`, `whoosh-fast`, `swoosh-text`, `pop-bubble`, `click-mouse`, `typing-short`, `ding-notify`, `boom-impact`, `riser-short`, `glitch-short`, `ka-ching`, `record-scratch`.
+- **Attribution required** for `cinematic-tense` and `meme-beat` (Kevin MacLeod, CC BY 4.0): paste the line from CREDITS.md into the post's description.
+- Known approximations: `corporate-light` is a sparse ambient track with short silent gaps; `tech-pulse` loops a 50 s original once (one audible seam); `glitch-short` is a digital "random" blip, not a long glitch.
+
+For more files, [`assets/audio/LIBRARIES.md`](../assets/audio/LIBRARIES.md) separates libraries whose files may ship inside a project from libraries that may only be **used** in a project (most "royalty-free" libraries forbid redistributing the file).
+
+## What the references use (confirmed by a source)
+
+Only what a creator's own description, credits or staff listing states; third-party "edit like X" tutorials do not count.
+
+| Creator | What is confirmed |
+|---|---|
+| Jeff Nippard | music from Epidemic Sound (video descriptions); a named sound editor |
+| Veritasium | music from Epidemic Sound (video descriptions) |
+| Kurzgesagt | original score and sound design by the studio Epic Mountain |
+| Nick DiGiovanni | a dedicated sound designer (staff listing); no library named |
+| Alex Hormozi | a sound-designer job posting; no library named |
+| Fireship, Greg Isenberg, NetworkChuck, Mark Tilbury, Drew Binsky, Filipe Deschamps, Porta dos Fundos, Niina Secrets | nothing found |
+
+Epidemic Sound is a subscription library: usable in a project under its licence, never redistributable in a package. The pattern behind the high-reach explainers is a **subscription library or a composer**, not free packs; the kit here is the free, shippable stand-in.
+
+## Levels
+
+Keep the bed well under the voice (about 0.2 of the voice volume in Remotion, faded in and out) and let `wizz-video-effects` duck it under speech and gate the master at its loudness target. Ad use of any track: check the licence column of `wizz-video-effects`' `sfx-licensing` rules (invoke the `wizz-video-effects` skill, references/sfx-licensing.md) before running paid.

package/src/skills-lib/wizz-video-styles/references/measured-patterns.md

@@ -0,0 +1,38 @@
+# Measured patterns across every style
+
+Set measured on 2026-10-04: **99 public shorts**, 46 in English and 53 in Brazilian Portuguese, 8 niche groups (9 to 16 per group), views from 23,718 to 107,324,401 (median 608,820). Per-style numbers live in [`data/styles.json`](../data/styles.json) and at the top of each card; this page keeps only what holds across styles. Method and limits: [research-method.md](research-method.md).
+
+## Sound effects are rare
+
+| | Cuts with a sound effect on them |
+|---|---|
+| All styles | 154 of 1,490 (10%) |
+| All styles except macro-asmr | 94 of 1,228 (8%) |
+
+The macro-asmr count is inflated by the music's kick drum landing on cuts. The rule that follows: **a sound on every cut is the exception, not the norm**. Budget sound for the two or three moments that matter; let `wizz-video-effects` enforce the ceiling.
+
+## Music bed depends on the style
+
+28 continuous, 36 faint, 35 none. Continuous beds cluster in the motion styles (motion-editorial 3/3, explicador-animado 5/6, meme-rapido 4/6); talking heads and podcast clips have none or a faint one (talking-head-legenda 0/21 continuous, podcast-clip 0/12).
+
+## Loudness
+
+Integrated loudness ranged from -29.0 to -8.9 LUFS, median -16.9. Pieces with a music bed sat around -15.5; voice-only pieces around -18.9. Platforms normalise playback, so master to the `wizz-video-effects` loudness gate (about -14 LUFS) rather than copying a reference's level.
+
+## Rhythm is a property of the style
+
+Median cuts per 10 s go from 0.45 (talking-head-cards) to 5.43 (esquete). Reach does not follow the cut rate: zero-cut single takes reached 600k (quadro-branco, BR health) and 5.4M (a stream clip), while one of the fastest styles (esquete) peaked at 2.9M. Pick the rhythm of the chosen style; do not speed up a style that is slow by nature.
+
+## Captions
+
+Read on contact sheets, not measured by a tool. Reach-heavy talking styles use 1 to 3 words at a time, bold, lower-centre, with one keyword in an accent colour (yellow is the BR default). Motion and explainer styles keep captions small or skip them; food, product and challenge styles often have none, because the image tells the story.
+
+## Hooks
+
+Read on contact sheets. The first frame already carries the claim: a headline box, a real number, the finished result, or the first stage of a challenge. None of the high-reach references opened on a logo or a greeting.
+
+## What could not be measured
+
+- The **name** of any sound effect or music track. What sources confirm is only the library or the people (see [audio.md](audio.md), "What the references use"). The catalogue gives sound **types** and their placement, and the kit in `assets/audio/` gives licensed equivalents.
+- Exact fonts (approximations from frames).
+- Retention: public data has views, not watch time.

package/src/skills-lib/wizz-video-styles/references/research-method.md

@@ -0,0 +1,33 @@
+# Research method, limits, and how to extend the catalogue
+
+## How a reference is measured
+
+`scripts/analyze.sh <youtube_id> <out_dir>` on a public YouTube Short:
+
+1. Downloads it at up to 720p with `yt-dlp` and reads views, likes, channel, date and title.
+2. **Cuts:** ffmpeg scene detection (`select='gt(scene,0.28)'`). Gives cuts per 10 s and the median scene length.
+3. **Audio** (Python with librosa + numpy): harmonic/percussive separation. The harmonic floor against its peak estimates a **music bed** (`continua` above 0.13, `fraca` above 0.07, else `sem`). Percussive onsets above the 80th percentile are classified by duration, low-frequency share, centroid and flatness into `pop_click`, `whoosh` or `impact_low`, and counted when they fall within 0.15 s of a cut.
+4. **Loudness:** ffmpeg `ebur128` integrated LUFS.
+5. **Contact sheet:** 16 frames across the video. Somebody must **read it** (open the image) before describing the look; the visual fields of a reference come only from that reading.
+6. Deletes the video file; keeps `<id>.json`, the sheet, the cut list and the metadata.
+
+Requires `yt-dlp`, `ffmpeg`, and a Python with librosa and numpy (`VIDEO_STYLES_PY=/path/to/python` when it is not the default `python3`).
+
+## Known limits
+
+- Scene detection undercounts cards and overlays sliding over the same shot, and counts nothing in a single take.
+- The SFX heuristic confuses speech plosives with whooshes in talking heads, and a music kick with impacts in beat-driven edits. Per-10-second SFX counts are therefore unreliable; the catalogue uses **SFX on cuts**, which is less noisy.
+- It never names a sound or a track.
+- Views are public views; there is no retention data.
+- Instagram and TikTok need a login and were not used; YouTube Shorts of the same creators stand in for them.
+
+## Adding a reference or a niche (any project)
+
+1. Find high-reach shorts: `yt-dlp "https://www.youtube.com/@CHANNEL/shorts" --flat-playlist --playlist-end 40 --print "%(view_count)s|%(id)s|%(title)s"`; for discovery, `yt-dlp "ytsearch30:<term> shorts" --flat-playlist --print "%(view_count)s|%(duration)s|%(id)s|%(channel)s|%(title)s"`. Prefer each channel's top views and vary the look.
+2. Measure each one, **one at a time**: `scripts/analyze.sh <id> research/<niche>/`.
+3. Open every contact sheet and write the visual fields.
+4. Append an object to `data/references.json` with the same keys as the existing ones (metrics copied from `<id>.json`, `style` set to an existing style key; a new style also needs a recipe in `data/recipes.json`).
+5. A new niche also needs a short key (one letter, the next free one) in `data/recipes.json["niches"]`, with its label; `nicho` in the reference uses that key. The build fails on an unregistered key.
+6. `python3 scripts/build_catalog.py` regenerates `data/styles.json` and every card. It refuses a style with no measured reference.
+
+Changes made inside a project live only in that project. To make them reach every project, contribute them to the Wizz Method repository (`src/skills-lib/wizz-video-styles/`): the next update installs them everywhere.

package/src/skills-lib/wizz-video-styles/references/style-picker.md

@@ -0,0 +1,51 @@
+# Style picker: material first, then niche and goal
+
+The material decides which styles are **possible**. Niche and goal only rank the possible ones. A style the material cannot feed is never offered, however well it performs.
+
+## 1. Material → possible styles
+
+| What the project has | Possible styles |
+|---|---|
+| Voice only (own recording or TTS), no one on camera | [motion-editorial](styles/motion-editorial.md), [explicador-animado](styles/explicador-animado.md), [meme-rapido](styles/meme-rapido.md) |
+| Voice + app/site screens | [motion-editorial](styles/motion-editorial.md) (screenshots in a phone card), [camera-no-setup](styles/camera-no-setup.md) (film the screen), [meme-rapido](styles/meme-rapido.md) |
+| A person on camera talking, nothing else | [talking-head-legenda](styles/talking-head-legenda.md) |
+| A person on camera + headlines, articles, charts, b-roll | [talking-head-cards](styles/talking-head-cards.md) (face bubble over full-screen b-roll is a variant) |
+| A recorded conversation, interview or talk | [podcast-clip](styles/podcast-clip.md) |
+| A camera filming the desk, phone in hand, hardware, a prop | [camera-no-setup](styles/camera-no-setup.md) |
+| A board, a notebook, a drawing app | [quadro-branco](styles/quadro-branco.md) |
+| A physical product or food filmed close, hands in frame | [macro-asmr](styles/macro-asmr.md) |
+| Something that can be done in stages, timed or priced | [desafio-numero](styles/desafio-numero.md) |
+| Handheld footage of a place or experience | [pov-vlog](styles/pov-vlog.md) |
+| A space to walk through (property, store, clinic) | [tour](styles/tour.md) |
+| A physical experiment or demonstration | [experimento](styles/experimento.md) |
+| Actors or one person in two roles, a script with a punchline | [esquete](styles/esquete.md) |
+
+**Mixed is normal.** The most common mix: the person on camera for the hook and the CTA, another style in the middle (motion editorial, cards, b-roll). Keep the face on screen as a bubble while b-roll is full screen; come back full screen for the CTA.
+
+## 2. Niche → styles seen there, by highest reach
+
+From [`data/references.json`](../data/references.json): the styles measured in each niche, ordered by the highest view count found. Reach depends on the channel as much as on the style; read this as "what works there", not as a ranking to obey.
+
+| Niche | Styles found (highest views) |
+|---|---|
+| dev / AI / build in public (EN) | meme-rapido (8.8M), camera-no-setup (8.1M), talking-head-legenda (5.4M), quadro-branco (4.4M), motion-editorial (0.83M), talking-head-cards (0.82M), podcast-clip (0.48M) |
+| dev / AI / tech business (BR) | quadro-branco (1.1M), talking-head-cards (0.41M), talking-head-legenda (0.26M), meme-rapido (0.23M), camera-no-setup (0.11M) |
+| personal finance / investing | podcast-clip (1.2M), talking-head-cards (1.2M), talking-head-legenda (0.77M) |
+| product / SaaS / marketing / business / law | desafio-numero (20.9M), talking-head-legenda (9.8M), camera-no-setup (0.53M), podcast-clip (0.33M), quadro-branco (0.15M), talking-head-cards (0.09M) |
+| health / fitness / nutrition | explicador-animado (33.4M), podcast-clip (9.8M), quadro-branco (0.61M), macro-asmr (0.54M), talking-head-cards (0.43M), desafio-numero (0.22M) |
+| food / recipes / lifestyle / travel | desafio-numero (107M), pov-vlog (8.7M), macro-asmr (7.7M), tour (1.05M) |
+| beauty / fashion / local services (clinic, salon, architecture, real estate) | macro-asmr (10.6M), talking-head-legenda (3.8M), talking-head-cards (2.0M), pov-vlog (1.4M), tour (0.37M) |
+| education / science / comedy | experimento (49M), explicador-animado (4.7M), esquete (2.9M), talking-head-cards (2.2M), talking-head-legenda (1.7M) |
+
+BR local services (clinic, salon, law, architecture), BR travel and BR sales/marketing were filled with a second pass (300k to 10.6M views). Still thin: BR software/SaaS proper (only sales/marketing found), real-estate agents in BR, and every niche has 9 to 16 references, not hundreds. When a niche is missing, pick by material and by the closest niche, and say the reference is borrowed.
+
+## 3. Goal → tie-breaker
+
+- **Saves and comments (teach, explain a system):** motion-editorial, explicador-animado, quadro-branco, talking-head-cards.
+- **Reach and shares (curiosity, humour):** desafio-numero, experimento, meme-rapido, esquete, pov-vlog.
+- **Trust and authority (a person the audience should know):** talking-head-legenda, podcast-clip, camera-no-setup.
+- **Desire for a product:** macro-asmr, desafio-numero, camera-no-setup.
+
+## 4. Write the decision down
+
+One line in the piece's plan: `estilo: <style> (+ <mix>) · porque: <material> + <niche> + <goal> · referências: <2 links from the card>`. Then open the card.

package/src/skills-lib/wizz-video-styles/references/styles/camera-no-setup.md

@@ -0,0 +1,60 @@
+# Camera filming the setup (Riley Brown, NetworkChuck, Theo)
+
+Measured on 6 shorts (en, pt) in: dev / AI / build in public (EN), dev / AI / tech business (BR), product / SaaS / marketing / business / law.
+
+## Measured
+
+| Metric | Value |
+|---|---|
+| Reach (views) | max 8,133,189 · median 283,565 |
+| Cuts per 10 s | min 0.35 · median 1.23 · max 2.89 |
+| Median scene length | 3.68 s |
+| Music bed (continuous / faint / none) | 2 / 1 / 3 |
+| Cuts with a sound effect on them | 2/79 (3%) |
+
+Cuts come from scene detection (it undercounts overlays and cards); music and SFX from an audio heuristic that classifies type, never names the sound ([../research-method.md](../research-method.md)).
+
+## Material it needs
+
+A real camera (or phone) filming the screen, the phone in hand, the hardware or a prop. Face optional.
+
+## When it fits
+
+Tool demos, AI coding, hardware, 'look what I built'. Feels authentic, not produced.
+
+## Recipe
+
+**Look**
+
+- Physical camera shots of monitor and phone with hands in frame; close-ups of the product; cuts to clean screen captures when detail is needed.
+- Headline at the top in the first seconds.
+
+**Palette:** Real room light, often warm or purple; captions white, keyword in a coloured box.
+
+**Typography:** Bold sans uppercase captions, 1 to 3 words.
+
+**Motion**
+
+- Handheld moves; a cut every ~4 to 8 s between face, hands, screen (NetworkChuck is the fast variant, ~2.7 s); punch-in on the screen detail.
+
+**Captions:** Short, uppercase, lower-centre, keyword in a coloured box (purple/yellow).
+
+**Hook (0 to 3 s):** The product or the result physically in hand in the first second.
+
+**CTA:** Spoken, sometimes 'link in bio' on screen.
+
+**Music:** Mixed: none or a continuous bed.
+
+**SFX:** Almost none on cuts.
+
+## Do not
+
+- Screen recordings that look stretched or tiny; film the screen or use a crisp capture.
+
+## Watch before editing
+
+- [NetworkChuck](https://www.youtube.com/shorts/tu1amrfflb8) · 8,133,189 views
+- [Seus Direitos](https://www.youtube.com/shorts/v69l3MZBn3g) · 528,901 views
+- [Theo - t3․gg](https://www.youtube.com/shorts/3uBiTsj8YSo) · 377,432 views
+- [Riley Brown](https://www.youtube.com/shorts/GZxjs8FD3Ck) · 189,698 views
+- [Riley Brown](https://www.youtube.com/shorts/pDsd2FiMUt8) · 151,630 views

package/src/skills-lib/wizz-video-styles/references/styles/desafio-numero.md

@@ -0,0 +1,60 @@
+# Challenge or comparison with a number on screen (Nick DiGiovanni, Mark Tilbury)
+
+Measured on 5 shorts (en, pt) in: product / SaaS / marketing / business / law, health / fitness / nutrition, food / recipes / lifestyle / travel.
+
+## Measured
+
+| Metric | Value |
+|---|---|
+| Reach (views) | max 107,324,401 · median 20,908,382 |
+| Cuts per 10 s | min 0.32 · median 3.35 · max 5.15 |
+| Median scene length | 2.3 s |
+| Music bed (continuous / faint / none) | 1 / 4 / 0 |
+| Cuts with a sound effect on them | 2/42 (5%) |
+
+Cuts come from scene detection (it undercounts overlays and cards); music and SFX from an audio heuristic that classifies type, never names the sound ([../research-method.md](../research-method.md)).
+
+## Material it needs
+
+Something to film doing the challenge: food, a product, a street interaction. A presenter is optional.
+
+## When it fits
+
+Food, money, products, everyday economics. The highest single reach in the set (107M).
+
+## Recipe
+
+**Look**
+
+- A visible progress device fixed on screen: time icons (1 s / 1 min / 1 h) changing colour per stage, emoji checklist, price tags on each ingredient.
+- Close hands, real places (market, street), a reaction at the end.
+
+**Palette:** Warm natural colours; tags in bright yellow with black outline; red/green icons for stages.
+
+**Typography:** Bold sans; price tags heavy with outline; often no captions at all.
+
+**Motion**
+
+- A cut per stage of the challenge (median scene ~2.3 s); ingredients drop in; tag pops on each price.
+
+**Captions:** Often none (the icons tell the story); when present, white bold word by word with the number bigger in yellow.
+
+**Hook (0 to 3 s):** The first stage already happening plus the device on screen (icon '1s', price '$9.84').
+
+**CTA:** The final reaction or the total; spoken follow.
+
+**Music:** Faint bed in most.
+
+**SFX:** Rare on cuts; a pop or tick when a tag or icon appears.
+
+## Do not
+
+- Hiding the number; the number is the story.
+
+## Watch before editing
+
+- [Nick DiGiovanni](https://www.youtube.com/shorts/wfTISOEGVqQ) · 107,324,401 views
+- [Joshua Weissman](https://www.youtube.com/shorts/lYhtV4zNyUg) · 31,976,392 views
+- [Mark Tilbury](https://www.youtube.com/shorts/qf5pfY5ljm8) · 20,908,382 views
+- [Mark Tilbury](https://www.youtube.com/shorts/mlCx2b404Rc) · 14,802,282 views
+- [Leandro Twin](https://www.youtube.com/shorts/CnwBXbBhC_E) · 217,387 views

package/src/skills-lib/wizz-video-styles/references/styles/esquete.md

@@ -0,0 +1,57 @@
+# Sketch (Porta dos Fundos, Manual do Mundo)
+
+Measured on 3 shorts (pt) in: education / science / comedy.
+
+## Measured
+
+| Metric | Value |
+|---|---|
+| Reach (views) | max 2,874,065 · median 2,247,836 |
+| Cuts per 10 s | min 3.77 · median 5.43 · max 5.66 |
+| Median scene length | 1.57 s |
+| Music bed (continuous / faint / none) | 0 / 1 / 2 |
+| Cuts with a sound effect on them | 5/134 (4%) |
+
+Cuts come from scene detection (it undercounts overlays and cards); music and SFX from an audio heuristic that classifies type, never names the sound ([../research-method.md](../research-method.md)).
+
+## Material it needs
+
+Actors (or one person in two roles), a set, a script with a punchline.
+
+## When it fits
+
+Comedy, branded content (the sketch carries the ad), education with humour.
+
+## Recipe
+
+**Look**
+
+- Fixed medium shots, cuts between speakers, inserts of hands/props; costumes change roles.
+
+**Palette:** Set colours; captions white bold with yellow/cyan keywords.
+
+**Typography:** Bold sans uppercase.
+
+**Motion**
+
+- Fast dialogue cuts (median ~5.4 per 10 s, scene ~1.6 s).
+
+**Captions:** Lower-centre, white bold, keywords coloured.
+
+**Hook (0 to 3 s):** The absurd situation in the first line of dialogue.
+
+**CTA:** The punchline; brand at the end if sponsored.
+
+**Music:** None or faint.
+
+**SFX:** Rare on cuts.
+
+## Do not
+
+- Setup longer than 3 s before the first joke.
+
+## Watch before editing
+
+- [Manual do Mundo](https://www.youtube.com/shorts/9iakv5VUYPs) · 2,874,065 views
+- [Porta dos Fundos](https://www.youtube.com/shorts/LTlTv4zChYo) · 2,247,836 views
+- [Porta dos Fundos](https://www.youtube.com/shorts/zKJSyVJB_yE) · 1,390,082 views

package/src/skills-lib/wizz-video-styles/references/styles/experimento.md

@@ -0,0 +1,56 @@
+# Experiment or demonstration (Veritasium)
+
+Measured on 2 shorts (en) in: education / science / comedy.
+
+## Measured
+
+| Metric | Value |
+|---|---|
+| Reach (views) | max 49,019,913 · median 30,514,714 |
+| Cuts per 10 s | min 1.79 · median 3.29 · max 4.78 |
+| Median scene length | 2.54 s |
+| Music bed (continuous / faint / none) | 1 / 1 / 0 |
+| Cuts with a sound effect on them | 0/43 (0%) |
+
+Cuts come from scene detection (it undercounts overlays and cards); music and SFX from an audio heuristic that classifies type, never names the sound ([../research-method.md](../research-method.md)).
+
+## Material it needs
+
+A physical experiment or a demonstration you can film, optionally street questions.
+
+## When it fits
+
+Science, education, product proof ('watch what happens when').
+
+## Recipe
+
+**Look**
+
+- A clean bench or real location; macro of the mechanism; montage between street poll and explanation.
+
+**Palette:** Natural; captions yellow with black outline or white with shadow.
+
+**Typography:** Bold sans with outline.
+
+**Motion**
+
+- Cuts every ~2.5 s between the setup, the reaction and the explanation.
+
+**Captions:** Lower-centre, yellow bold with outline, key words highlighted.
+
+**Hook (0 to 3 s):** The question over the close-up of the experiment ('what if you cut the green rope?').
+
+**CTA:** The reveal is the payoff; spoken follow.
+
+**Music:** Faint or continuous, unobtrusive.
+
+**SFX:** None measured on cuts.
+
+## Do not
+
+- Revealing the answer before the question lands.
+
+## Watch before editing
+
+- [Veritasium](https://www.youtube.com/shorts/itSkBESLZeY) · 49,019,913 views
+- [Veritasium](https://www.youtube.com/shorts/2GUah9xHVto) · 12,009,516 views

package/src/skills-lib/wizz-video-styles/references/styles/explicador-animado.md

@@ -0,0 +1,62 @@
+# Animated explainer (Kurzgesagt, Jeff Nippard, Jeremy Ethier)
+
+Measured on 6 shorts (en) in: health / fitness / nutrition, education / science / comedy.
+
+## Measured
+
+| Metric | Value |
+|---|---|
+| Reach (views) | max 33,439,880 · median 17,692,306 |
+| Cuts per 10 s | min 1.81 · median 2.79 · max 5.25 |
+| Median scene length | 2.48 s |
+| Music bed (continuous / faint / none) | 5 / 1 / 0 |
+| Cuts with a sound effect on them | 18/109 (17%) |
+
+Cuts come from scene detection (it undercounts overlays and cards); music and SFX from an audio heuristic that classifies type, never names the sound ([../research-method.md](../research-method.md)).
+
+## Material it needs
+
+A voice track and animation (2D illustration, 3D anatomy, infographics, charts). Face optional, often as b-roll.
+
+## When it fits
+
+Science, health, fitness, nutrition, finance concepts, any 'how it works'. Among the highest median reach of every style measured.
+
+## Recipe
+
+**Look**
+
+- Flat saturated illustration (Kurzgesagt) or a dark gym / light neutral background with 3D anatomy and highlights.
+- Red or warm glow over the subject (the muscle, the organ, the number); dashed arrows; before/after comparisons; charts that build.
+
+**Palette:** Saturated illustration palette, or grey/charcoal base with one red highlight and one green/neon secondary.
+
+**Typography:** Heavy italic sans for titles; plain sans for captions; thin serif or sans for data labels.
+
+**Motion**
+
+- Scene change every 2 to 3 s; elements build in sequence; highlight pulses on the subject; charts draw.
+- Camera moves inside the illustration rather than cutting.
+
+**Captions:** Small, white with outline, lower-centre, often no colour emphasis; titles big in a gradient or bold colour.
+
+**Hook (0 to 3 s):** The claim plus the visual of the subject highlighted, or a data point ('19% body fat') in the first second.
+
+**CTA:** A summary table or the full routine/sheet on screen; spoken follow/save.
+
+**Music:** Continuous bed in almost every case; the richest scores of the set (Kurzgesagt's are composed by Epic Mountain).
+
+**SFX:** Whooshes on transitions are common here, still only on a minority of cuts; soft UI sounds on highlights.
+
+## Do not
+
+- Static slides with voice over.
+- More than one highlight colour fighting for attention.
+
+## Watch before editing
+
+- [Jeremy Ethier](https://www.youtube.com/shorts/u9OQMBPrFgI) · 33,439,880 views
+- [Jeff Nippard](https://www.youtube.com/shorts/1mGXae-LCiM) · 24,154,920 views
+- [Jeff Nippard](https://www.youtube.com/shorts/t9Cs6HXy1v8) · 23,593,044 views
+- [Jeremy Ethier](https://www.youtube.com/shorts/zG2ux0MQRhM) · 11,791,568 views
+- [Kurzgesagt – In a Nutshell](https://www.youtube.com/shorts/2cK8l5Yg5w8) · 4,743,841 views

package/src/skills-lib/wizz-video-styles/references/styles/macro-asmr.md

@@ -0,0 +1,61 @@
+# Close-up product or recipe (Niina Secrets, Receitas de Pai, Ana Maria Brogui)
+
+Measured on 11 shorts (en, pt) in: health / fitness / nutrition, food / recipes / lifestyle / travel, beauty / fashion / local services (clinic, salon, architecture, real estate).
+
+## Measured
+
+| Metric | Value |
+|---|---|
+| Reach (views) | max 10,597,557 · median 541,087 |
+| Cuts per 10 s | min 0.0 · median 3.87 · max 9.86 |
+| Median scene length | 2.15 s |
+| Music bed (continuous / faint / none) | 3 / 5 / 3 |
+| Cuts with a sound effect on them | 60/262 (23%) |
+
+Cuts come from scene detection (it undercounts overlays and cards); music and SFX from an audio heuristic that classifies type, never names the sound ([../research-method.md](../research-method.md)).
+
+## Material it needs
+
+A physical product or food filmed close (overhead or macro), hands in frame. Face optional.
+
+## When it fits
+
+Beauty, food, crafts, any physical product. Fast and tactile.
+
+## Recipe
+
+**Look**
+
+- Macro of texture (gloss, sauce, dough), overhead table with ingredients, hands working; the finished result shown in the first second.
+- Natural sound of the process (sizzle, pour, click) is part of the piece.
+
+**Palette:** The product's own colours on a neutral surface (beige, wood, checked cloth).
+
+**Typography:** Little or no text; a watermark or logo; ingredient labels in bold if any.
+
+**Motion**
+
+- Fast cutting: median ~3.9 cuts per 10 s (scene ~2.2 s), up to ~10 in product macros; each cut is a step.
+
+**Captions:** Usually none; quantities in a box if it is a recipe.
+
+**Hook (0 to 3 s):** The final result (bite, swatch, finished dish) in the first second.
+
+**CTA:** Logo card or spoken; the recipe in the caption/description.
+
+**Music:** Mixed; a beat in beauty, faint or none in recipes where the process sound matters.
+
+**SFX:** Measured hits here are partly the music's kick; treat the number as uncertain. Process sounds matter more than added SFX.
+
+## Do not
+
+- Covering the product with text.
+- Slow steps; every step is one shot.
+
+## Watch before editing
+
+- [NiinaSecrets](https://www.youtube.com/shorts/geZQMV1N5ak) · 10,597,557 views
+- [Joshua Weissman](https://www.youtube.com/shorts/2KyDrqiiZdU) · 7,705,698 views
+- [NiinaSecrets](https://www.youtube.com/shorts/OF_U-Xjd-B4) · 5,456,546 views
+- [Doutora Acne](https://www.youtube.com/shorts/_XEiyCA7xvc) · 2,880,265 views
+- [Receitas de Pai](https://www.youtube.com/shorts/041o6T400ZA) · 2,226,212 views

package/src/skills-lib/wizz-video-styles/references/styles/meme-rapido.md

@@ -0,0 +1,65 @@
+# Fast meme montage (Fireship, Filipe Deschamps)
+
+Measured on 6 shorts (en, pt) in: dev / AI / build in public (EN), dev / AI / tech business (BR).
+
+## Measured
+
+| Metric | Value |
+|---|---|
+| Reach (views) | max 8,779,355 · median 1,810,275 |
+| Cuts per 10 s | min 0.19 · median 3.5 · max 5.25 |
+| Median scene length | 2.42 s |
+| Music bed (continuous / faint / none) | 4 / 2 / 0 |
+| Cuts with a sound effect on them | 10/138 (7%) |
+
+Cuts come from scene detection (it undercounts overlays and cards); music and SFX from an audio heuristic that classifies type, never names the sound ([../research-method.md](../research-method.md)).
+
+## Material it needs
+
+A voice track and a stock of images: code, memes, stock clips, logos, article screenshots. Face optional.
+
+## When it fits
+
+Dev and tech news, explainers with humour, history of a technology, opinion pieces. Highest reach per dev channel in the set.
+
+## Recipe
+
+**Look**
+
+- Dark or saturated flat background (black, navy, electric blue).
+- Big display text, code on screen, stock footage, meme reaction clips, hand-drawn stick figures, logos stacking on a timeline.
+- Highlighter pills on key words (yellow/orange on BR channels).
+
+**Palette:** Black or navy base; 2 to 4 loud accents (yellow, cyan, lime, orange, purple). Official logo colours when logos appear.
+
+**Typography:** Heavy rounded or pixel display font for key words; mono for code; bold sans for pills.
+
+**Motion**
+
+- Hard cuts every 1.5 to 3 s; text pops in; stills slide or scale briefly.
+- Memes and reaction clips as punchlines, usually 1 to 2 s.
+- Timeline variant: items accumulate on one canvas, almost no cuts.
+
+**Captions:** Key words only, very large and coloured, centred; or a highlighter pill of 2 to 4 lowercase words lower-centre (BR variant).
+
+**Hook (0 to 3 s):** A surprising fact or number in huge type in the first second.
+
+**CTA:** Usually spoken; a final meme or the channel's signature card.
+
+**Music:** Continuous bed in most (upbeat lo-fi / electronic), low under the voice.
+
+**SFX:** Whoosh on some cuts and a pop on text, but only on a minority of cuts (see the measured row).
+
+## Do not
+
+- A sound on every cut.
+- Memes longer than the joke (more than ~2 s).
+- Small text: this style lives on huge type.
+
+## Watch before editing
+
+- [Fireship](https://www.youtube.com/shorts/s9F8pu5KfyM) · 8,779,355 views
+- [Fireship](https://www.youtube.com/shorts/aXcuz6fn8_w) · 5,220,458 views
+- [Fireship](https://www.youtube.com/shorts/8B20fRB78nA) · 3,388,073 views
+- [Filipe Deschamps](https://www.youtube.com/shorts/vusWY-1EGLk) · 232,477 views
+- [Filipe Deschamps](https://www.youtube.com/shorts/i_sd01c2kbw) · 176,999 views