Packet 0: Bible review, 2026-09-04

Objective

Improve three decisions on the Manufactured Love bible before Phase 3 spends money on reference images:

1. Roster: which of the 8 crew, 6 cast, 1 host to cut, merge, rename, or keep. A 120 s trailer can introduce a limited number of faces.
2. Lead: Marcus Bell (senior story producer) as lead with Kaya Morales (PA) as the last face before the title, versus Dana Kessler (showrunner) as lead, versus another shape.
3. Wardrobe and attractiveness: Daniel's direction, verbatim: "Make sure some of the characters are HOT AS FUCK and dressed accordingly as appropriate for the character." Decide which characters carry the sex appeal, how each is dressed so it reads on a 768p AI frame, and what stays locked.

Source map (read these; do not trust this packet's summary over the files)

• bible/show.md: series, Sunburn format, Bulu brand, Look A / Look B, never-do list.
• bible/characters.md: 15 character blocks with wardrobe hex, props, never-change lists, voice, signature line, sin, arc.
• docs/PLAN.md: acceptance criteria (A1 to A7), locked decisions D1 to D7, phases, $500 cap, cost basis.
• docs/research/R1-production-reality.md and glossary.md: sourced production reality; every manipulation in the script must map to a term here.
• docs/research/R2-genre-grammar.md: style sheet (Look A villa, Look B truth), confessional setups, wardrobe conventions of the genre.
• docs/research/R3-trailer-grammar.md Part C: the 120 s beat sheet with timecodes and per-section shot lengths (cast run is 0:08 to 0:30, freeze frame + lower third per producer).
• docs/research/R5-longform-practice.md sections 1, 6, 7: reference-locking rules, per-model reference limits (Wan 3.0 Prime 10 images, H3 Max 12 files, Seedance 2.5 up to 50), prompting for realistic humans, per-shot cost.
• docs/research/unreal-differentiation.md: rules that keep the show distinct from UnREAL.

Hard rails (from PLAN.md and show.md, not negotiable in this review)

• Dark comedy, deadpan. No contestant ever speaks to the documentary camera.
• Manipulation, alcohol, sex: dialogue and implied visuals only. Nothing shown outright.
• Bulu parody trade dress, no real platform marks. 16:9 only. $500 total generation cap.
• Every trick in the trailer maps to a sourced R1 technique.

Current hypothesis (challenge it)

• Eight crew is at the ceiling of what 22 s of cast run can introduce; the beat sheet gives roughly 1.5 to 2.5 s per freeze-frame introduction, so 8 introductions eat the whole setup section.
• Marcus as lead because he is the most quotable and voices The Shade, which gives the trailer its cruelest audio device. Kaya as the last face because a PA topping up a glass is the cleanest implied-not-shown image.
• Sex appeal currently sits with the contestants only (coral bikini, white linen, lilac set). The crew is dressed as crew. That may be wrong for the genre and for the audience: the producers are the stars of this show, and reality-TV crews on a villa shoot are young, tanned, and dressed for heat.

Known gaps

• No reference images exist yet; nobody has seen a face.
• Wardrobe hex values were chosen for distinctness across characters, not checked against each other or against Look B's low-light interiors.
• No table read has happened; signature lines are untested.
• Attractiveness is unspecified for every crew character.

Risk map

• Spend: Phase 3 reference sheets cost 10 to 20 image generations per character. Every character kept costs money; every character cut after Phase 3 wastes it.
• Safety filters: swimwear and "hot" wardrobe on crew may trip Seedance 2.5 moderation; Wan and H3 Max are more permissive per R4/R5 but unverified.
• Likeness: no character may resemble a real named producer, exec, or host.
• Representation: the roster should read as a real 2026 UK/US unscripted crew, not a checklist and not a monoculture.

Per-lens questions

Each lens answers all three decisions and the attractiveness question, but leads with its own.

• Unscripted showrunner / trailer editor lens: how many faces can 0:08 to 0:30 introduce; who earns a freeze frame; who is the lead; does the trailer plot in show.md section 1 land in 120 s; which signature lines are actually funny on a table read.
• Casting and costume lens: who is hot, how hot, and dressed how, per character, with the genre's own wardrobe language (R2) and 2026 villa-crew realism (R1 physical production); make every "hot" choice specific to the character's job, age, and status; propose exact garments, fit, skin shown, hair, grooming, with hex colors; flag anything that would read as costume rather than clothing.
• AI generation feasibility lens: can each look be held across 50 shots at 768p; are the never-change lists the right anchors (glasses, hair, one garment, one prop); which wardrobe colors collide with each other or with Look B interiors; which props will fail (radios, earpieces, clipboards, sticky notes); how many characters per shot the reference limits allow; which characters could share a light sheet; what "hot" wardrobe risks filter refusals per model and how to phrase it.
• Skeptical simplifier / comedy writer lens: what to delete; is the PA arc a cliché and how to make it not one; is Marcus or Dana the lead; are the jokes procedural (actions) or written (quips); does anything drift back toward UnREAL; which lines to cut or rewrite; the one image the trailer is remembered for.

Output contract

## Verdict (three decisions plus attractiveness, one line each)
## Sources / Files Inspected
## Roster table: character | keep / cut / merge / rename | why | hot? (none / attractive / very / lead sex appeal) | wardrobe change (exact garments, hex) | never-change list edits
## Lead recommendation and why
## Strongest Objection to the hypothesis
## Lines to cut or rewrite (quote, replacement)
## Risks I would flag to Daniel
## P0 / P1 / P2 Action Queue
## What Would Change My Mind
## Confidence

Do not modify any file. Read-only review. Quote file paths and line numbers for evidence. Under 120 lines.

Launch ledger

Acceptance proof for the review

Four reports in the contract above. A synthesis in docs/reviews/2026-09-04-bible-review.md with decisions taken, and the bible files updated to v0.2 with a changelog.