Courtroom — sources

Reference ads (Foreplay, analysed 10-06 in ~/research/davidaistar/foreplay/)

• 10_character_arguments — Penrose Skin scented body lotion, 62 s, 9:16, AI 3D "podcast debate". Beats (paraphrased): a jar-headed lotion character boasts it outlasts a glass cologne-bottle character → the bottle answers with heritage and a high price → lotion says women prefer him and the bottle buys attention → the bottle defends its concentration, lotion says alcohol evaporates in hours → a woman steps in and praises the lotion → lotion explains its mechanism (butter locks the scent in), the bottle lunges and gets shattered → price comparison and CTA. Tells: no true lip-sync, warped background faces, AI physics on the shatter; EL voices. Solution/product-aware, comparison angle. Keep: two characters trading short jabs, a third party who settles it, the mechanism delivered as an argument point, the product character winning on the facts. Change: our dispute is about HER (problem-aware, recipe A), not product vs product; staged as a hearing so the third party has real authority; no violence or shatter gag; no double-entendre; the band stays dark and calm (silent product, silent winner). Gemini's suggested Dawn concepts ("Screaming Clock vs Silent Ninja", "Ignored Phone vs the Band") are the seed for our objects cut, rebuilt so the stake still lands on Mom.
• 04_drama — Koriderm Time Reverse Cream, 4:32, 9:16. Beats: a clerk rejects her ID → a supervisor compares the face → an interrogation listing procedures she supposedly had → she trades her secret for the stamp → a hand test of the product → timeline + guarantee → the stamp, she leaves vindicated. ~95% AI, lip-synced over generated video; tells: gibberish text on the ID, morphing label, fused fingers on application. Keep: an official process as the stage, the protagonist accused in front of an authority, the outcome delivered by the authority. Change: no lip-sync (core §3), so lines are staged on listeners; no product application by hand; the official outcome has to be one that office can really give (realism table).
• Exhibit A (our song, sa-ontime-10-exhibita-a / ad-04-exhibit-a-v2, disney3d). Lyrics in output/concept-packs/dawnbands-song-reskin-exhibita-clay-2026-10-06.md. Beats: the ex tells the mediator she doesn't care about their son's education, tardies as proof → her mornings (knocks, the bus, late to her own job) → "Exhibit A", the printout, "Is this accurate?" → two-week trial with the dad → her low point in the car → app pings: late, absent, late → the school nurse's root cause (smoke-alarm study, sound doesn't reach him) → she finds and tries the band → the kid's photo from the bus stop on the follow-up morning → the ex's printout now carries his two weeks → the kid clears her and holds out his wrist → close + 60 nights. Keep: the whole recipe A spine. Change: told live in the room, not sung in hindsight; the kid's words reach the mediator realistically (she can ask him; in court it's the judge's summary).

davidaistar method notes (analysis/)

• diff_podcast_dialogue.md: "80% of realism is the script" → short alternating turns, no monologues (turn-taking rules). His on-camera talk comes from one native-voice generation per speaker; ours is staged until a native-speech mode exists (core §3). He hides weak lip-sync with masks and distractions; we stage the listener instead.
• diff_animation_styles.md: anthropomorphic characters as a format family (mascot explains the fix vs our villain-personifies-the-problem); one model per style; cross-batch continuity is not built (run.py builds refs per batch), so the hearing room must come from one env sheet to look the same across batches.
• diff_copy.md: no dialogue-script doctrine exists in skills/direct-response-copywriter/references/formats.md; this skill's turn-taking rules are the dialogue doctrine for this format until that sub-mode is written.
• REPORT.md §3 row 10/11 ("High as a courtroom version, exhibit device = best song CPA 2.48 ROAS") and §4 #4 (mom vs accuser, alarm as exhibit A, band as the witness).

Our receipts

• MAP.md §2A device row: exhibit 2.48 vs experiment 1.67 vs testimonial 1.26; vindication payoff $1,638 at 1.81; accuser close to her; legal/family stakes landing on mom.
• song-ad-playbook.md §1: exhibit A $33.7 CPA; §2 real-life catches (mediator "ordering" custody); §5 scene types (evidence insert, witness/escalation, vindication with arm held OUT).
• dawn-stakes.md: #3 "only allowed 18 absences before she reaches truancy status" [reddit 1t3hejh]; #6 "threatening letters… You can get put in jail" [reddit 1rfciee]; #7 "they might involve cps" [reddit 199olsy]; CPS/custody-neglect lines with no morning tie are excluded.
• 10-01 S2 attendance-office packs: the attendance office as a set, the clerk as explainer, the dialogue pace keys, voices MOM/CLERK. Its intents still say "mouth moving" (written before the core §3 rule); don't copy that staging.
• Code checked 10-06: pack.py (speaker regex, product-word flag), run.py ~L158-170 (dialogue cfg; style_* crash reproduced with a scratch pack), vo.py (per-line takes, gap_turn/gap_same, tempo_/style_ per speaker), vo_check.py (DIALOGUE limits, EXEMPT names), storyboard.py (SHORT_BEAT_S 1.2 merges within a line only; panel writer prompt "mouth moving"), h3.py ("No on-camera dialogue, no lip-sync"), sheet.py, endcard.py, scripts/statics/voc_check.py.
• cartoon-pipeline: villain_parade / talking_ingredients = Kling 3.0 native dialogue via Higgsfield (use_elevenlabs_vo=False), one speaker per scene, intro template; cp-014 Kling native ≤2.8 wps + one bucket + 0.5 s tail.

starpop articles (creator's own write-ups, read 2026-10-06; ~/research/davidaistar/starpop/txt/)

• Applied: facebook-instagram-ad-sizes-and-safe-zones + how-to-preview-a-facebook-ad-before-publishing → §10 placement gate (the article names the UI regions but gives no percentages; the 14/35/6 % Reels box is Meta's published spec, cross-checked 10-06). are-ai-generated-ads-allowed-on-meta → §11 upload hygiene (Advantage+ enhancements off, live preview day 1). meta-video-ad-metrics-hook-rate-hold-rate-ctr-cpa, how-much-budget-to-test-meta-ad-creatives → §11 read/floor lines where this format needed them.
• Read, nothing new for this format: ai-ad-formats-guide (format ↔ job table, shot-list order, references before video: already core §4), fully-ai-generated-ads-are-here-to-stay (strategy essay), how-to-calculate-break-even-roas-for-meta-ads (proposed for core §6, not format-specific), how-to-add-correct-captions-in-any-language + how-to-fix-wrong-capcut-auto-captions (script-as-source captions = our assemble rule already; his SRT → CapCut/Premiere import maps to Remotion per core §2a).
• Not adopted: (1) "an AI actor must not voice a fictional first-person experience / fabricated testimonial" (formats guide, AI-ads article, organic-vs-paid) — conflicts with Fish's claims unlock (10-02) and our dramatized mom-narrator recipe; risk is Fish's call, recorded here only. (2) His 25-char vertical caption line cap — VideoAd chunks by its own MAX_PHRASE_CHARS 28 + two-line fit, Fish-set. (3) Organic-first discovery (ai-organic-social-vs-paid-ads) — Dawn's message and checkout path are proven, Testing CBO is our discovery lane. (4) His model picks (Kling O3 Pro, Sora 2, P-Video) — registry roles only; untested candidates.