← Files PixVerseARCHIVED FILE

skills/pixverse-video-script/references/tech.md

2.38 KB · Oct 4, 2026 · 12:28 UTC

↓ Download file

# Tech Review / Product Test / Tutorial

Apply `./genre-routing.md` first. Resolve product, format, actual work performed, honest
verdict, untested scope and runtime. A hands-on hook needs truthful effort plus a supported
number; absent testing uses a sourced-analysis/tutorial lane without fabricated experience.
Write backward from the verdict. Source pace is160–200English words/min on camera,
200–235for dense VO; adapt for clarity, available footage and requested language.

Three hooks, each≤3sentences/15s: actual test ledger, fresh evidence, verdict with a deeper
why, real failure, access obstacle, countdown stake, contrarian test or payoff montage.
Do not create fake refusals, purchases or benchmarks to realize these engines.

Single review: hook/evidence0–5%, what-it-is5–12%, walkthrough12–70%, numbered downsides
70–85%, verdict+who-it-suits+untested-scope85–95%, one-line ending. Tour in curiosity,
physical or usage order, never spec-sheet order. Show the strongest feature early and
plant specific later-feature deferrals only when the runtime supports them.
Ranking: announce sort rule in sentence2; each item setup→demo→score/verdict, maintain
a real running tally and final total. News: recency→short context→findings with per-item
verdict/caveat→final calibration. Tutorial: show actual task outcome, then ordered steps.

Translate every spec into an experience, comparison or exact overlay. Demo before judgment;
skepticism is the baseline, excitement follows evidence. Own opinions; every downside has
a fairness beat and who-it-still-suits/what-would-fix-it. Hype needs a comparison class.
Distinguish future uncertainty, small sample and unverified inference. A long explanation
lands on a short verdict sentence; introduce an analogy promptly for a hard concept.

Long work may defer one named feature per~2min, pivot per section and hold one major
total/rank for final20%. Pay every named deferral and close the actual hook. No vague
“later,” mid-test monetization, press-release prose or unearned superlatives.
At≤4min: verdict hook0–10%, three proof beats strongest first10–70%, one catch70–85%,
who-it-suits plus scope then stop. No feature-deferral stack; at most one held number.
English3min roughly500–580words; timestamp15–30s, retention3–5rows. A vertical product
does something in first2s. Deliver three hooks, complete script, demo/overlay cues and map.

SHA-256: 4207709bc0a2d076a8605049d884c35bcbe769b7640fa9f832fa41713076a453