Explainer / Skill-Tips Reels
A sister style to the documentary look: faster, card-driven, built to teach
what each tool does in 45–65 s. Shares the brand config, voice rules, export
and QA of vertical-reels-video-studio-pro.
First run
If brand.config.json is missing, has "configured": false or still says @yourhandle, run step 0 of
vertical-reels-video-studio-pro first: ask the user for their social handles (IG/Threads/TikTok/X), watermark text and
position, CTA text + handle, and voice, and save them in brand.config.json. Never default to anyone else's handle.
What makes a good tips Reel
- Exact names on screen. Show the real
owner/repo(e.g.remotion-dev/skills,FFmpeg/FFmpeg) large and spelled exactly. Double-check spelling in the captions too (a caption "ffmpag" is a bug). - One job per tool. For each item say what it does in one plain sentence and show a matching, accurate example (its README hero, a real output, a short real code excerpt). No vague tool lists.
- No fabricated proof. Never invent stars, downloads, prices, results or "made with" claims. If a reference video was not made with every listed repo, don't say it was.
- Primary sources. Read the repo README, licence and official docs. Credit the original post you are reacting to, and keep the original author separate from the authors of the tools (the person who posted a list did not write the repos in it).
- Free vs paid. Keep official public skills distinct from paid/private skill packs. Never promise to give away someone else's paid product.
Script shape (casual, direct, first person)
| Beat | Time | On screen | VO |
|---|---|---|---|
| Hook | 0–3 s | Huge hook + the reference clip or result | Bold claim or question |
| What it is | 3–10 s | Repo card owner/repo, licence chip | One sentence |
| Items 1–N | ~8 s each | 2–3 readable cards per beat, never all at once | What EACH does + example |
| How to use | ~8 s | Real command / prompt in a mono card | The one step that matters |
| Check | ~5 s | Phone-size crop, "readable? cut off?" | Quality tip |
| CTA | last 4–6 s | Big CTA card + repo line, presenter gesture | the user's {{CTA}} (asked at first run), e.g. "Want all the skills? DM me." |
The CTA invites the viewer to act. You never send DMs or post automatically.
Layout
- House palette and fonts (see
docs/STYLE_GUIDE.md). Cards:#14202Bfill,#384A582 px border, 24 px radius, hard drop shadow0 12px 0 #05090E. - Kinetic stressed words, floating readable cards, small breadcrumb per scene
(
HOOK · WHAT · HOW · CHECK · CTA), restrained transitions. - A full, balanced 9:16 frame with no dead footer, while respecting the Reels safe zone (content y 200–1500; captions ~y1310–1353; the user's watermark ~y1427–1446 if bottom; everything left of x≈960).
- Landscape reference footage: contain it full-width with deliberate negative space, or alternate full view and one readable close-up. Never stretch it and never blindly centre-crop away UI or credits. A clip from the reference is a contained example, not the theme of the whole video.
- Source credit line on every scene that shows someone else's work
(
Tool · Author · owner/repo · Licence).
Presenter character
- Ask the user whether they have their own presenter character; never reuse someone else's.
- One consistent identity (same face, outfit, accessories) in every video.
- Body + arm on a shoulder pivot (independent sleeve and hand, fixed torso and head) for an up/down "teaching" gesture; a whole-image bob alone looks fake. Check for detached joints and clipping at the extreme angles.
- Mouth from a VO-driven track (
scripts/build-mouth-track.py). - Pointing/"presenting" poses must face the thing being explained.
Voice
- Same voice in every video; pronunciation rules from
docs/VOICEOVER.md(A.I., tickers spelled out, repo names readable — re-voice a line if the repo name is unintelligible). Hard command names stay on screen, not in VO. - Audit actual pronunciation, captions, timing and duplicates; re-voice only the failing lines.
Workflow
- Draft + fact-check first (
templates/draft-template.md). Obey "don't render". - After approval: TTS → timing → build → render (GPU if available; report device).
- Export (
scripts/reels-export.sh), preview < 40 MB, brief end hold (~0.5 s). - Mandatory audio + visual review, then deliver covers A–D, captions,
.srtand source credits.