Skip to content

Likeness — Startup Design Output

Mode: Fast Track end-to-end with Deep research depth Date completed: 2026-05-09 Output root: design/ (at repo root)

Executive Summary

Likeness is a consent-first AI likeness platform for verified adult creators. Verified creators license a private AI version of themselves to subscribers within rules they set, watermarked, revocable, with no model export. The platform sits at a structurally favorable intersection of the adult creator subscription market ($7.22B OnlyFans GMV 2024) and the adult AI image generation market ($1-3B in 2025) — with a regulatory direction (TAKE IT DOWN Act enacted; NO FAKES Act pending; California AB 2602; Mastercard adult content rules) that increasingly mandates the exact consent infrastructure Likeness's architecture commits to.

The competitive cell — consent-first + explicit-allowed + verified-creator-licensed + strong architectural commitments (no model export, license-gated inference, per-creator isolation) — is uncontested today. The closest funded competitor, Vylit ($2.7M Seed, ex-OnlyFans CEO Ami Gan), explicitly excludes explicit content. The window is real but finite (12-18 months estimated).

The pre-seed plan ($1.5M for 18 months, six-seat founding team, US-first) is appropriately sized. The biggest risks are concentrated in three places: payment processor stability (Civitai precedent), creator participation validation (zero structured interviews to date), and the speed of the Creator Ops cofounder hire (the founder is outside-in to the adult industry; this hire bridges the founder-market-fit gap).

Verdict: 6.7/10 — CONDITIONAL PROCEED. Three validations gate deeper commitment. Concrete experiments and kill criteria are in 06-validation/validation-playbook.md.

Key Findings

From research

  • Adult AI is mass-market — 337+ apps generating ~$1B+ collectively in 2025 — but no platform serves real verified creators on a consent-first basis. SoulGen, Promptchan, Candy.ai, DreamGF use synthetic personas. Civitai (LoRA marketplace) was de-banked by their processor in May 2025 over nonconsensual content.
  • Fans demonstrably pay for AI on creator platforms. 15% of Fanvue revenue is AI-generated content (Sacra). This is the strongest single demand signal.
  • Vylit is the most acute direct competitor. $2.7M Seed (Sept 2025), launched early 2026, founded by ex-OnlyFans CEO. Same thesis as Likeness but explicitly excludes explicit content. Strong founder-market fit. Likeness's positioning advantage is the explicit-content scope where most adult monetization lives.
  • Regulatory direction is favorable. TAKE IT DOWN Act (signed May 2025) requires 48-hour platform takedown. NO FAKES Act (pending) provides a federal right of publicity that's "licensable but not assignable" — directly aligned with Likeness's revocation premise. California AB 2602 (Sept 2024) voids non-specific digital replica contracts.
  • OnlyFans's own AI policy (verified-only, with disclosure) soft-validates Likeness's positioning. OnlyFans won't build the AI tooling itself due to Mastercard/Visa exposure.

From strategy synthesis

  • Likeness should not position as an OnlyFans replacement. Position as an additive AI revenue layer that creators' existing fans plug into.
  • Architectural commitments are the real positioning — the consent infrastructure that competitors retrofitting under regulatory pressure can't match cheaply.
  • C2PA membership pre-launch is a high-leverage low-cost credibility move.

Top 3 Risks and Mitigations

Risk Mitigation
Payment processor de-banks (Civitai precedent) Multi-processor redundancy from day 1; CEO-led BD pre-launch; consent posture documented and pitched as a processor-relationship asset
Creator participation fails to materialize (no primary discovery to date) Front-load Experiment 1 (10-15 structured creator interviews) before deep capital commitment beyond pre-seed; specific kill criterion at ≤3 of 10 saying yes
Vylit or Fanvue ships a directly competitive product within 12 months Speed of Creator Ops cofounder hire is competitively load-bearing; differentiation on architectural commitments; quarterly competitive monitoring

Confidence Dashboard Summary

  • High confidence: Regulatory landscape; macro market sizing (OnlyFans 2024 financials, Fanvue ARR); Vylit / Loti / Vermillio funding facts; Mastercard rules.
  • Medium confidence: Creator economics benchmarks; processor fee ranges; pricing tier WTP intuition.
  • Low confidence — known unknown: Creator participation in Likeness's specific configuration. Cannot be filled by web research; requires structured interviews. This is the load-bearing unknown.

Full details: 01-discovery/confidence-dashboard.md.

Anti-Patterns Detected

Reviewed for common founder anti-patterns:

  • Solution looking for a problem: ❌ NO. The problem (unauthorized AI use of creators) is well-documented and growing.
  • Boiling the ocean: ❌ NO. MVP scope is appropriately narrow (5-10 invite-only creators, image-only, US-first, no video/voice/self-insert).
  • Premature scaling: ❌ NO. Concierge phase is explicitly first; closed beta is post-validation.
  • Vanity metrics: ❌ NO. Key metrics (02-strategy/lean-canvas.md) are AARRR-focused with operational floor metrics.
  • Building in stealth too long: ⚠ POTENTIAL. The mockup is built but no creators have seen it. Validation playbook directly addresses this.
  • Ignoring unit economics: ❌ NO. Unit economics are explicitly modeled (05-financial/revenue-model.md) with sensitivity analysis identifying the leveraged variable.

Document Index

00-intake/

  • brief.md — Synthesized intake from existing repo docs + founder Q&A
  • brainstorm.md — Three brainstorm variations; convergence with absorbed lessons

01-discovery/

Synthesized deliverables: - market-analysis.md — TAM/SAM/SOM, growth, regulatory summary, geographic, timing - competitor-landscape.md — Comparison matrix, positioning map, vulnerability analysis - target-audience.md — Persona, pain hierarchy, jobs-to-be-done, language map (with discovery-gap caveats) - industry-trends.md — Regulatory trajectory, technology shifts, investment signals - confidence-dashboard.md — Per-claim confidence with source-tier and corroboration counts - verification-report.md — Self-verification of synthesized deliverables - research-gate.md — Yellow Light go/no-go assessment with named conditions

Raw findings (in raw/): market-size, regulatory, trends, adjacent-markets, competitors, customers-demand, distribution

02-strategy/

  • lean-canvas.md — Complete Lean Canvas (Fast Track scope)

03-brand/

  • mission-vision-values.md — Three mission options, three vision options, recommended values set
  • tone-of-voice.md — Personality traits, voice principles, writing samples by surface, vocabulary guide
  • brand-personality.md — Sage / Outlaw / Caregiver archetype mix, emotional targets, visual direction, competitive contrast

04-product/

  • mvp-definition.md — Core hypothesis, 30 must-have features, scope discipline, success criteria
  • feature-prioritization.md — RICE-scored build order in three phases (months 1-3 / 3-6 / 6-9)
  • user-journey.md — Two-persona journey with aha moments and drop-off risk analysis

05-financial/

  • revenue-model.md — Pricing strategy, revenue mix, unit economics anchors, sensitivity analysis, key assumptions to test

06-validation/

  • validation-playbook.md — Top 3 experiments + risk highlights + kill criteria
  • scorecard.md — Final 6.7/10 score with verdict and breakdown

Project root

  • PROGRESS.md — Phase tracker
  • README.md — This file
  • action-plan-30-days.md — Concrete weekly plan for the first month post-funding

Phases done in full (originally skipped under Fast Track, expanded by user request)

  • Phase 5 — Brand: Three deliverables in 03-brand/. Treats existing CLAUDE.md and mockup as POCs; produces fresh work the founder can compare against and choose to adopt, iterate, or merge.
  • Phase 6 — Product: Three deliverables in 04-product/. Formalizes MVP definition, RICE-prioritized feature list, and end-to-end user journeys for both creator and fan.

Files NOT generated (per Fast Track scope)

The standard startup-design skill produces additional files: expanded financial projections (cost-structure, projections), expanded strategy (value-proposition, business-model, positioning, go-to-market as separate files), and expanded validation (risk-analysis, assumptions-tracker, experiment-design, kill-criteria as separate files). Under Fast Track, those are folded into the produced files or omitted because the existing repo docs cover them. Phase 5 and Phase 6 were originally skipped under Fast Track but expanded in this session.

Next Step

Read action-plan-30-days.md. The first month's work is dominated by two things: closing pre-seed and starting Experiment 1 (creator discovery). Everything else either supports those or waits.