tasyakamila.com — Product Plan (REBUILD)

AI clone platform — multi-era persona of Tasya Kamila. Interactive AI chat + AI vocal generation across four life eras (Kids 1998-2004, Teen 2005-2009, Young Adult 2010-2018, Present 2019-now). Fans pick the era they want to talk to — or hear sing. Architecturally agnostic ("the platform"); branded with Tasya Kamila as the flagship character (user handles licensing).

1. The wedge — why this platform wins

Three converging shifts in 2026 make this category newly live:

The empty shape nobody owns: multi-era AI personas of established Indonesian celebrities. Closest analogues:

• Character.AI — English-first, generic IP, no era selector
• Replika — generic companion, no real-celebrity personas
• HeyGen / Synthesia — corporate avatars, no celebrity-IP licensing layer
• Suno — vocal generation but no persona / no era / no chat
• Indonesian celebrity Linktrees — zero AI interactivity

We fill it: persistent multi-era AI of one well-known Indonesian celebrity, voice-cloned across her actual career timeline, interactive in chat AND able to sing new songs.

2. Money stack (5 tiers, IDR + USD)

Year-1 ARR target: Rp 9B ($560K). Break-even by month 8 with ~3K paid users.

3. Acquisition (triple-stack)

1. TikTok "Talk to Kids Tasya" trend — 30-sec videos of fans showing their Kids-Tasya chat or song output. The AI clone IS the viral hook. Post 5x/week on official + creator-driven.
2. WhatsApp shareable songs — generated songs export to WA-friendly format with "Made on tasyakamila.com" footer. Every song = a free billboard. Same playbook 18music.net would use.
3. Twibi-Twibi reactivation — original Cherrybelle / Tasya fanbase from 2011-era nostalgia, now in their 20s-30s, are the highest-LTV cohort. Email + IG re-engagement campaign at launch.

4. AI automation map

Per chat msg total cost: ~$0.005 (sustainable at Free tier). Per song total cost: ~$0.16 (Fan tier breakeven at 3-4 songs / month — pricing supports it).

5. Cloudflare architecture

• Worker: tasyakamila-com — SSR landing + era selector + auth gate
• Worker: tasyakamila-chat — chat API + persona routing + memory
• Worker: tasyakamila-sing — song generation queue + voice synthesis
• Pages: tasyakamila-app — React app for chat + song studio UI (post-Phase-2)
• D1: tasyakamila_db — accounts, chat history, songs, subscriptions, era preferences, audit ledger (EU AI Act compliance)
• R2: tasyakamila-audio — generated songs (encrypted at rest, signed URLs)
• R2: tasyakamila-reference — training reference audio per era (label-only access)
• KV: tasyakamila-cache — era-prompt cache, voice-id cache
• Durable Objects: ChatSession — per-user conversation state across the 4 eras
• Queue: tasyakamila-render-q — song generation fan-out (Suno + ElevenLabs)
• Cron: weekly token-usage report + monthly compliance audit + daily memory snapshot

6. 30 / 90 day plan

Day 1-2 (now): Landing + era selector + booking-of-licenses funnel + 9 sub-pages Day 3-14: D1 schema + Stripe + Xendit checkout + free-tier signup flow Day 15-28: Chat MVP — 4 era personas wired to Claude Sonnet 4.6 with hand-crafted system prompts per era. NO voice yet. Day 29-45: Voice clone — train 4 ElevenLabs PVCs (Kids / Teen / Young Adult / Present) on reference audio Day 46-60: Sing MVP — Suno v4 + ElevenLabs overlay, user picks era voice + genre + theme Day 61-75: EU AI Act compliance layer — C2PA signature + audio watermark + audit log + disclosure overlay Day 76-90: First brand license signed + Founders 100 limited drop + public launch

7. Moat

• Multi-era voice library compounds with use — each subsequent month of refinement on PVC models makes them harder to replicate from scratch
• Real-celebrity IP license (user handles legal side) = entry barrier nobody else has for Tasya specifically
• EU AI Act Article 50 compliance baked in = enterprise brand-license sales advantage starting Aug 2026
• Persistent memory per user = switching cost — fans build months of chat history that doesn't port to other AI companions
• WA-share viral loop + Indonesian-language voice quality = nobody else will reach SEA mom-pop market at our cost basis

8. Mandatory standards

• 100% Cloudflare. Mobile-native landing + chat UI + song studio. PWA.
• EU AI Act compliance native — every chat message labeled "AI Tasya · [era]"; every generated song carries C2PA signature + visible "AI Avatar" disclosure overlay + embedded audio watermark
• NO real child photos in marketing — kids-era visuals are atmospheric (90s TV, cassettes, classroom scenes) implying the era without depicting kids
• Per-era persona system prompts documented and version-controlled — auditable
• Trademark + likeness license handled by user (we build platform-agnostic, ship branded for Tasya as flagship)
• WCAG AA on every text color

9. Competitor synthesis

10. Missing parts audit

• Domain + CF zone live
• Licensing contract — Tasya likeness + voice rights (user handles, Phase-1 blocker before voice training begins)
• ElevenLabs Pro account + 4× Professional Voice Clone slots (Rp 1.5jt/month — Phase 3)
• Reference audio archive per era (1+ min per era minimum for PVC) — Phase 3
• Suno API key OR Udio fallback — Phase 4
• Stripe + Xendit Indonesia merchant — Phase 2
• Privacy / Terms / EU AI Act disclosure templates reviewed by Indonesian + EU counsel — Phase 1
• C2PA + AudioSeal toolchain on Cloudflare Workers — Phase 5
• Persona system prompt library (4 era prompts, version-controlled, A/B-tested with fan reactions) — Phase 3
• Founders 100 drop mechanism + early-access flag — Phase 6

⚙ HARD CONSTRAINTS (enforced for all sites)

This domain MUST operate within these constraints — no exceptions:

• 100% Cloudflare serverless — Workers + D1 + R2 + KV + Workers AI + Vectorize. NEVER PM2, NEVER VPS, NEVER Docker in production path.
• 100% AI-automated — every customer interaction, every moderation decision, every transaction reconcile = AI. No manual queue, no live human chat support, no physical fulfillment.
• 1-operator solo — one person can run the entire operation from a phone. No team meetings, no shared inbox, no shift rotation.
• WhatsApp AI bot for all support (24/7, instant response, no SLA promises that need humans).
• Mayar QRIS for all Indonesian payments (subscription auto-renew, no manual invoicing).
• Indonesian UI primary — bahasa-first, English fallback only where unavoidable.
• Privacy — opt-in only, delete-on-request honored within 24h (cron-driven).
• No physical goods, no inventory — digital products + affiliate referrals only.

If the plan above describes any flow that violates these constraints, treat the plan as ASPIRATIONAL only and rework before building. The constraint trifecta wins.