← back to the board · all wave-1 PRDs

PRD CL08 · UKIYO TEST WAVE 1

Status: Ready to make as a local composite ($0) once the review is chosen and UKIYO confirms its service.

1. The ad

Ad CL08 · static 4:5
ad_id mint before production — no 16-pilot/mint-CL08.json on disk
Objective Conversion: WhatsApp booking of the RM198 first trial, judged on cost per booked trial from the chat log + the per-ad code, never on cost per chat alone (map-v4 objective).
Offer line 首次细胞光体验 RM198
Type · pain testimonial · pain B: skin that reddens; worry before any treatment
ICP Same catchment; skin that reddens easily and a worry that a treatment will make it worse (ICP-MAP.md §0 point 2: 'redness-prone skin', 1.86, within noise of the top segment).
Hook (map-v4) 「客人的真实回复」 + 「脸容易泛红的话」
Structure from map-v4 CL08: a real guest's WhatsApp message with a new redness pain line and RM198, on the layout of your S2 ad (06-winning-dna/evidence/ukiyo-S2.jpg).

2. References — what is copied, what never is

Ref Source (file · record) Signal We copy Never copy
UKIYO's own S2 guest-message ad 06-winning-dna/evidence/ukiyo-S2.jpg; DNA.md lines 453 and 578 (A3) ad 52534646961850 · deepest chats in the account The layout: 「客人的真实回复」 title, 效果因人而异, the real chat bubble, offer line, logo. its RM99; the review-62 bubble (organic-only, result claim 皮肤隔天很亮); quotes from another treatment under the CellLight offer (DNA A3: proof must be the treatment sold)
Dr Stan AD46 patient-review card 02-drstan-winners/WINNERS.md §'#8 DRSTAN-MH-AD46' 4 bookings · disapproved by Meta Let a real person's own words carry the pain-to-relief. star rating and the 4.9★ aggregate; a named reviewer / honorific; the 70% stat; the AD46 file itself (kept in its lane only, brand firewall)
Your ref 5 · SKIN HOUSE case cards (IG Da13mqYH0MW) 11-operator-refs/REFS.md §'Ref 5' MY chain · 3-card carousel · 4 likes (unproven) Condition label → real matched before/after → the disclosure; one condition per card. their photos and quotes (third-party clients); their treatment codes; any generated before/after

3. The static

Layout S2's layout, 4:5 on the UKIYO cream paper ground: small kicker 「脸容易泛红的话」 → big title 「客人的真实回复」 → 效果因人而异 → the guest's real WhatsApp bubble (screenshot pasted as is, never retyped or edited) → offer line → WhatsApp 预约 → logo.
Every text line 「脸容易泛红的话」 · 「客人的真实回复」 · 「效果因人而异」 · 「首次细胞光体验 RM198」 · 「WhatsApp 预约」
Picture Author pick: ukiyo-review-21 (workspace/ukiyo/testimonial/ukiyo-review-21.md on origin/master 4f41d7e5e7; usable_in organic + paid; consent confirmed-by-ukiyo; one clean bubble 「满意啊 护理师很温柔 手法也纯熟」 12:22 PM; no identifiers). It answers the worry (gentle hands) and passes the scanner. Alternate: ukiyo-review-36 「满意的 全程都很详细 很细心」 (passes; its screenshot also shows UKIYO's own quoted message). Not used: ukiyo-review-09, the S2 bubble about a sensitive nose — closest to redness, paid-cleared, but her own words carry 放松 / 舒服 / 不舒服, which the scanner and your relax/feel rule block (cl08-quote-check.scan.json).
Device panel None on this card (proof = the guest's words).
Proof A real guest's own words, paid-cleared.
Engine Local composite (PIL) of the real screenshot on the approved cream plate; no generation, $0.
Finish intel/ukiyo-reskin-statics-2026-09-24/finish.py with GRAIN_AMP=0.12, logo bottom-centre (real wordmark via smart-logo-placement).

4. Engine and cost

Paid video seconds 0 s of Wan 3.0 i2v
Est. USD (one fire per shot) $0.00 at 1080p ($0.00 at 720p); $0.00 if every shot needs one retry
Voice take none
Keyframes / statics codex-image-gen or local composite, $0 (subscription / local)

5. Compliance

legal_scan.py --selftest PASS, then legal_scan.py 24-wave1-prd/wave1-lines.json: CL08 pass (hook + 5 on-screen + 0 voice lines). Hits: none. Review flags: none.

Extra code check for the scanner's known blind spots (17-dpl/COMPLIANCE.md §3) and this run's rules (放松 / 舒缓 / 健康 / Korean claims / eye shields): none found.

Guest-quote check (24-wave1-prd/cl08-quote-check.json): review09_full: FAIL ["BANNED '放松' (relax-selling)", "BANNED '舒服' (feel line (F4 open))"]; review21: pass; review36: pass.

Rules held (where they apply to this ad): real pain first; results only as how skin looks + 效果因人而异; offer line only (no visit details); no relax / lie-back / feel lines; no eye-shield mention; no 'you have X'; generated people never pose as customers; before/after only from the real UKIYO client; device = the cream paddle with the red light-bar, never the purple-logo head; machine keeps its own 'Cell Light III Platinum' print; aesthetician at the head of the bed.

6. Blocking items

7. Steps to launch

  1. Clear the blocking items above (mint first).
  2. Build the static per the section above; finish (grain 0.12 + real logo bottom-centre); look gate.
  3. qa-doctor G7 (target 9.5) and legal_scan on the final text.
  4. Your approval; /approvals row; Meta upload with ad_name = the ad_id and its tracked_url; per-ad code in the WhatsApp prefill; primary text + headline legal-scanned; G8 published-atom.

Every Wan clip and every final lands on chain at action time: S3 copy, then a G5 video/canvas atom parented to this ad's visual-evidence atom, then the /approvals row (CLAUDE.md forward-emission rule).

← back to the board · all wave-1 PRDs