Skip to content

gooseworks-ai/goose-skills / skills/video/write-video-ad-script

write-video-ad-script

Verified

Write or revise a short-form ad from a sourced angle and reference. Plan connected words, product proof and sound; expose production gaps before script review.

references/reference-and-performance.md

View on GitHub
# Reference, proof and speech performance

Preserve the requested mechanism across words, action and sound. Complete source
observation establishes what happens; current brand facts support adapted claims.
The existing script review handles choices and gaps; this adds no approval round.

## Reference-to-script comparison

Inspect complete moving picture and audible sound with a capable reviewer. Frames
and transcription support observation, not listening or continuous-motion proof.
Record source identity/checksum, duration, covered ranges, methods and missing duties.
An audio attachment being returned does not establish hearing.

For each important beat save source interval/observation, adapted words, required
visible action/proof, speech/sound, planned scene/route and permitted change. Separate
the source's observed test from the user's adapted request. Resolve disputed mappings
before using a reference as authority.

A live demonstration, creator story or routine cannot silently become cards/dots or
generic beauty footage. List necessary action order and complete payoff. Use exact
product/variant views and sourced dimensions/use/capacity where material; do not guess
scale from name/weight or call generated illustration an independently measured test.
Resolve unsupported action with a concrete alternative in the current review.

Derive picture/direction from one scene intent; contradictions cannot clear script
review. Preserve exact user copy. Material hook/payoff, action/proof, sound or product
changes refresh affected existing review, not merely changes to spoken words.

## Speech and sound

Record speech separately from soundtrack. Native dialogue, narration, exact-driver
lip-sync, singing, timed mixed intervals and none have different requirements.
Music/ambience/effects remain separate; no detected words cannot infer total silence.

Non-speaking coverage can use accepted narration. Flexible-voice speaking creators
can use qualified native speech; exact supplied performance needs a qualified
audio-driven lip-sync route. Voice-reference timbre is not waveform reuse. Keep the
same performance under B-roll and source-link a returning speaking face.

The script names no voice. When an on-screen person speaks or is narrated, the voice is
paired with that person after they are made (create-creator-takes-h3's `voice_pair`, from
their age, gender and ethnicity), so leave every voice id out of the script and its draft.

Keep exact words separate from supported energy/emotion/emphasis/pause direction.
Save pronunciations for names, ingredients, numbers and abbreviations beside real
caption spelling. Use only the selected model/wrapper's supported tags/settings;
literal hesitation/laughter is spoken content only when intended. Do not decorate
every sentence or inherit a different model's speed/style controls.

Write connected thought/exchange with natural sentence-length variation. A short
reply can be human; repeated disconnected spoken headings are not. Brisk ordinary
UGC is a starting intent; the actual reference and accepted performance decide pace.
Calm proof holds and intentional pauses remain valid. Caption chunks are display
units, not mandatory sentence lengths.

Timing estimates stay provisional until accepted speech exists. Do not pad every
clause or automatically speed voice to rescue a poor timeline. Hear whole auditions,
joins and endings; watch relevant mouths. Missing listening remains incomplete;
scores/transcripts cannot certify pronunciation, believable voice or fine sync.
receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementationStars: 322kUpdated Claude CodeCodexVerifiedInstall
requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirementsStars: 322kUpdated Claude CodeCodexVerifiedInstall
using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallbackStars: 322kUpdated Claude CodeCodexVerifiedInstall
verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions alwaysStars: 322kUpdated Claude CodeCodexVerifiedInstall