From Idea to Launch

22 months of dedicated engineering — from first concept, through a Telegram stealth launch that grew to 3,000+ users with zero ads, to the public release of a full-featured AI NSFW platform.

Q4 2024 — Ongoing

Q4 2024: Research & Concept

October — December 2024

Market Research & Vision

Completed Extensive analysis of the AI content generation landscape, competitor platforms, and user needs. Defined the vision for a privacy-first, anonymous AI content platform that puts users in control.

  • Competitive landscape analysis of 20+ platforms
  • User persona research and target audience definition
  • Privacy-first architecture decision (minimal data collection, no email required)
  • Anonymous payment strategy (crypto + prepaid cards)
  • Platform naming, branding & visual identity design

Technology Stack Selection

Completed Evaluated dozens of AI frameworks, backend stacks, and frontend libraries. Selected the optimal combination for performance, scalability, and developer productivity.

  • FastAPI (Python) chosen for async backend performance
  • React 18 + TypeScript + Vite for modern frontend
  • Tailwind CSS + Radix UI for design system
  • Zustand for lightweight state management
  • SQLAlchemy 2.0 async ORM for database layer
  • ComfyUI selected as AI inference engine

Stable Diffusion Model Research

Completed Tested and evaluated hundreds of Stable Diffusion checkpoints, LoRA models, and inference configurations to find the optimal setup for photorealistic NSFW content generation.

  • Evaluated 50+ SD checkpoints for photorealism
  • RealisticVision v5.1 selected as primary model
  • Tested SDXL vs SD 1.5 performance trade-offs
  • Optimal resolution: 768×1152 (SD 1.5) for GPU compatibility
  • LoRA compatibility testing across model families
  • Inference benchmarking (81 sec/image at 35 steps on GTX 1080)

Q1 2025: Prototyping & Core Architecture

January — March 2025

ComfyUI Integration Engine

Completed Built a robust bridge between our platform and ComfyUI, enabling programmatic workflow execution with real-time progress tracking via WebSocket connections.

  • ComfyUI API client with async HTTP + WebSocket
  • Workflow template system (JSON-based)
  • Real-time generation progress via WebSocket streaming
  • Queue management and interrupt capabilities
  • Error handling and automatic reconnection
  • Image proxy with in-memory caching (100 images)

Database Architecture

Completed Designed and implemented a comprehensive database schema with 50+ models, supporting everything from user management to complex content relationships and financial transactions.

  • 50+ database models with full relationship mapping
  • SQLite with WAL mode for concurrent access
  • Async session management with aiosqlite
  • Foreign key constraints and cascading deletes
  • Automatic migrations on server startup
  • PostgreSQL-ready schema for production scaling

First Working Prototype

Completed End-to-end proof of concept: user uploads a prompt, ComfyUI generates an image, result displayed in the browser. The seed that grew into the full platform.

  • Text-to-image generation working end-to-end
  • Basic web UI with prompt input and result display
  • ComfyUI workflow execution pipeline proven
  • Image storage and retrieval system
  • Foundation for all future AI tools

Q2 2025: AI Tools Development

April — June 2025

AI Video Generation

Completed WAN 2.1/2.2 video generation integration — transform still images into animated video clips using state-of-the-art diffusion video models.

  • Image-to-video generation with WAN 2.2 models
  • High-noise and low-noise inference paths
  • LightX2V 4-step LoRA for fast generation
  • CLIP Vision + UMT5-XXL text encoders
  • FP8 quantized models for memory efficiency

Q3 2025: LoRA System & Content Engine

July — September 2025

LoRA Model Infrastructure

Completed Built a comprehensive LoRA management system — cataloguing, categorizing, and intelligently combining 141 specialized LoRA models for diverse content generation.

  • 141 LoRA models catalogued from disk with full metadata
  • 6 categories: Poses (52), Body Features (27), Utility (13), Special (7), Style (6)
  • Conflict group detection (breast size, pubic style, body type)
  • 14 preset combinations for common content types
  • Model-LoRA many-to-many assignment system
  • Automatic conflict resolution when combining LoRAs
  • LoRA strength and weight optimization per category

NSFW Content Category System

Completed Designed and implemented 26 distinct NSFW content categories with intelligent prompt engineering, category-specific LoRA chains, and model-content mapping.

  • 26 content categories: poses, close-ups, body features, settings, solo, toys, special
  • Category-to-LoRA mapping with automatic chain building
  • Intelligent prompt builder with category-specific keywords and fragments
  • Model-category mapping with special category restrictions
  • POSE_PROMPT_FRAGMENTS with detailed scene descriptions
  • POV framing for 35 interactive categories

NSFW Image Generation Service

Completed Full orchestration service combining FaceID consistency, LoRA chains, prompt building, and quality verification into a single automated pipeline.

  • End-to-end NSFW generation orchestration
  • IP-Adapter FaceID Plus v2 for face consistency across generations
  • Batch photo generation (30+ images per session)
  • ComfyUI workflow execution with WebSocket progress
  • Face embedding storage and retrieval

Prompt Engineering System

Completed Hardened prompt templates with safety guardrails, mandatory positive/negative terms, and category-specific keyword databases ensuring consistent, high-quality output.

  • Central prompt builder service (prompt_constants.py)
  • MANDATORY_POSITIVE_PREFIX: (1girl:1.3), (female:1.2), (woman:1.2)
  • MANDATORY_NEGATIVE_TERMS: male exclusion list
  • Category-specific keyword fragments and scene descriptions
  • Hardened templates (not user-configurable) for safety
  • Consistent prompt structure across all generation modes

Q4 2025: Security, Auth & Server Architecture

October — December 2025

Authentication & Privacy System

Completed Built a privacy-first authentication system — users can create accounts with just a code (no email required), or stay anonymous-friendly with free credits.

  • JWT-based authentication with secure token management
  • Anonymous code-based login — no email or personal data required
  • Traditional email + password registration (optional)
  • Session management with token refresh mechanism
  • Password recovery flow
  • Account deletion with full data purge
  • Frontend auth via Zustand persisted store

Security Infrastructure

Completed Enterprise-grade security layers protecting every endpoint — rate limiting, CORS, CSRF, input validation, and XSS prevention baked into the core.

  • Rate limiting on all API endpoints (security/rate_limit.py)
  • CORS restriction to allowed domains only
  • CSRF middleware protection
  • Security headers (HSTS, CSP, X-Frame-Options)
  • Input validation on all file uploads
  • XSS prevention (DOMPurify + React escaping)
  • Open redirect prevention on payment callbacks
  • Parameterized queries preventing SQL injection

Modular Server Architecture

Completed Refactored from a monolithic 2,852-line server.py into 60+ modular routers with clean domain separation, dependency injection, and structured error handling.

  • Split into 60+ modular FastAPI APIRouters
  • Domain-driven router organization (auth, profiles, feeds, payments, admin)
  • CurrentUser dependency injection for auth
  • Structured error responses with exception handlers
  • Health check and ComfyUI status endpoints
  • Dynamic bot rendering for SEO (pre-rendered HTML with JSON-LD)

SEO & Internationalization

Completed Full i18n support for 5 languages, XML sitemaps, RSS feeds, structured data, and dynamic bot rendering — every category page optimized for search engines.

  • 5 languages: English, German, Spanish, French, Italian
  • 14 i18n namespaces with complete translations
  • XML sitemap index with sub-sitemaps (1,360+ URLs)
  • hrefLang alternates for all locales
  • RSS 2.0 feed for all categories
  • OrganizationSchema + WebSiteSchema + ServiceSchema (JSON-LD)
  • Bot User-Agent detection with pre-rendered HTML
  • llms.txt and llms-full.txt for AI discoverability

Q1 2026: Full Platform Build & Polish

January — March 2026

200+ AI Model Profiles

Completed Created a diverse library of 200+ unique AI model profiles, each with distinct appearance, personality, backstory, and visual consistency through assigned LoRA models.

  • 200+ unique AI model profiles with full backstories
  • 8 ethnicities: Caucasian, Asian, Latina, African, Middle Eastern, Mixed, Indian, Pacific Islander
  • 10 personality types: Playful, Mysterious, Dominant, Submissive, Romantic, Athletic, Intellectual, Wild, Sweet, Edgy
  • 6 body types: Slim, Athletic, Curvy, Petite, Thick, BBW
  • 4 breast sizes: Small, Medium, Large, XL
  • SEO-friendly @username slugs for every model
  • Model-specific LoRA assignments for visual consistency
  • Admin management UI with grid, filtering, and tabbed editor

Social Feed System

Completed Built a complete OnlyFans-style feed experience with chronological, following, and algorithmic "For You" feeds — powered by a hybrid recommendation engine.

  • 3 feed types: Chronological, Following, For You (algorithmic)
  • Hybrid recommendation engine (engagement + recency + diversity)
  • 7 post types: Photo, Video, Text, Poll, Album, Story, Pinned
  • Cursor-based pagination with infinite scroll
  • IntersectionObserver-based automatic content loading
  • Poll system with custom options, voting, and percentage display
  • Rich media support with unified content schema

Social Interaction Layer

Completed Full social experience — likes, threaded comments, follows, AI auto-replies with personality matching, and real-time notifications via Server-Sent Events.

  • Like/unlike with atomic counters and optimistic UI
  • Threaded comment system with nested replies and pagination
  • Follow/unfollow with follower counts and suggestions
  • AI auto-reply matching each model's personality type
  • Configurable reply delay (0–300 seconds) per model
  • Real-time SSE notification streaming (6 notification types)
  • Mark as read, notification history, auto-reconnect on disconnect
  • Virtual gifts and gift tracking per post

Discovery & Search Engine

Completed Powerful discovery system combining grid browsing, advanced filtering, full-text search, Tinder-style swiping, and AI-powered recommendations.

  • Model grid with thumbnails, stats, and instant preview
  • Advanced multi-filter system: ethnicity, age range, body type, personality, content category
  • Full-text search (FTS5) with instant results and suggestions
  • Tinder-style swipe interface (react-tinder-card)
  • AI recommendation engine (319 lines, engagement-based)
  • Trending content tracking and explore page
  • Tag cloud browsing with category tabs
  • Filter count indicators and reset functionality

Monetization & Payment System

Completed Complete anonymous payment infrastructure — credit wallets, cards, wallets and crypto via a hosted checkout, subscription tiers, PPV unlocks, tips with threshold rewards, and creator payouts.

  • Hosted checkout: cards, Apple Pay, Google Pay and crypto
  • Anonymous payment options — no email required
  • Credit wallet system with transaction history (13 transaction types)
  • Pay-per-view content unlocks with subscription discounts
  • Tip system with 4 threshold reward levels (100 / 500 / 1,000 / 2,500 credits)
  • Anonymous tipping option (excluded from leaderboards)
  • Top tipper leaderboards with progress tracking
  • Subscription tiers per model (Basic, Premium)
  • Creator earnings aggregation (PPV + Tips + Subscriptions)
  • Payout processing with fraud detection and velocity checks
  • Payment webhook handlers with signature verification

270+ Content Categories

Completed Massive content library with 270+ unique categories spanning body types, ethnicities, poses, settings, and styles — each with multilingual FAQs, gallery images, and SEO landing pages.

  • 270+ unique content categories with full metadata
  • Categories by type: poses, close-ups, body features, pubic styles, settings, solo, toys, special
  • i18n FAQs in 5 languages per category (EN, ES, FR, DE, IT)
  • 5+ gallery images per category (WebP optimized)
  • Category-specific hero images for landing pages
  • SEO-optimized NicheLandingTemplate for every category
  • Keyword database for prompt generation (category_keywords.py)
  • NicheCTA with related category suggestions

AI Chat & Direct Messaging

Completed Private 1-on-1 conversations with AI models — personality-aware responses, chat credits, message history, and real-time delivery.

  • Direct messaging with any AI model
  • Personality-matched AI responses (10 personality types)
  • Chat credit system (separate from generation credits)
  • Full message history and conversation threads
  • Real-time message delivery notifications
  • Chat pricing page with credit bundles

Academy Educational Platform

Completed Learn AI image generation with structured courses — video and text lessons, progression tracking, instructor profiles, and membership access.

  • Multi-lesson course curriculum (4–15 courses)
  • Video and text lesson types with intro previews
  • Course progression tracking per user
  • Academy membership system with access control
  • Instructor profiles and course thumbnails
  • DOMPurify-sanitized lesson content (XSS-safe)
  • Lesson text verification and validation pipeline

Creator Economy Tools

Completed Full creator toolkit — onboarding flow, dashboard with analytics, post creation, subscription tier management, earnings tracking, and payout configuration.

  • Creator onboarding flow with step-by-step setup
  • Creator dashboard with KPI overview
  • Rich post creation (photo, video, text, poll, album)
  • Subscription tier management (create, edit, price)
  • Earnings dashboard (PPV, tips, subscriptions breakdown)
  • Payout settings and processing configuration
  • Model profile editor with tabbed interface

Admin Control Center

Completed Comprehensive admin dashboard — user management, content moderation, analytics, audit logs, promo/deal management, scheduler, and QC oversight.

  • 13 admin pages: Dashboard, Users, Creators, Models, Moderation, Analytics, Payouts, Promos, Deals, Audit Log, Settings, Scheduler, QC
  • Real-time analytics with KPI cards and charts
  • User and content moderation with flagging system
  • Audit log with timestamp, user, and action tracking
  • Promo code and deal management system
  • Background task scheduler (payouts, activity, generation)
  • AdminRoute protection with role verification
  • AdminLayout with dedicated navigation

Quality Control & Image Integrity Pipeline

Completed AI-powered 3-stage QC pipeline ensuring every generated image meets quality standards — gender detection rejects male faces, face verification ensures model consistency, and PIL checks validate technical quality.

  • 3-stage QC: Gender pre-gate → Face pre-gate → PIL quality checks
  • InsightFace gender detection (buffalo_l model, CPU)
  • Male-only rejection logic (female and unknown accepted)
  • Face verification with cosine similarity scoring (0.6 standard, 0.7 avatar)
  • Auto-retry: max 3 attempts per image with seed rotation
  • Per-attempt persistence (attempt_number, seed_used)
  • Permanent rejection marking for failed images
  • WebSocket substatus updates during QC pipeline
  • Photo visibility filtering (is_visible convention)

Strict Image Assignment Pipeline

Completed Ensures every image is assigned to the correct AI model — face-similarity-driven ranking, URL-category scoring, unique constraints, and admin review queue.

  • Face-similarity-driven image-to-model assignment (primary ranking)
  • URL-category scoring as tiebreaker
  • Unique constraint: one image per model assignment
  • ModelImageAssignmentDB junction table
  • Cleanup script with --dry-run safety flag
  • Admin QC metrics dashboard with per-model pass/fail rates
  • Admin review queue for rejected images with category filtering
  • Approve/delete actions with rejection reason color coding

Security Hardening & Testing

Completed Comprehensive pre-launch security audit — auth hardening on all endpoints, 361+ backend tests, E2E testing with Playwright, error handling at every layer, and dead code cleanup.

  • 361+ backend unit tests with pytest
  • Auth hardening on all protected routers
  • E2E testing with Playwright (critical user flows)
  • Comprehensive error handling at all layers
  • Pydantic model validation on all request/response schemas
  • Dead code cleanup and import optimization
  • TypeScript strict mode in frontend
  • Component testing with Vitest + React Testing Library

Content Completion & Final Polish

Completed Final quality pass — regenerating category images with inline QC, ensuring all 270+ categories have 5+ verified images with zero male faces, and polishing the generation pipeline.

  • Category image regeneration with inline QC pipeline
  • All 270+ categories with 5+ QC-approved WebP images
  • Zero male faces across all category galleries (InsightFace verified)
  • 108 male-detected images identified and replaced
  • Generation pipeline optimization (config, prompts, quality settings)
  • Frontend gallery synchronization (update_categories_gallery.py)
  • Final UI/UX polish and responsive design verification
  • 71 categories requiring image top-up in --fill mode

STEALTH: Stealth Launch via Telegram

April 1, 2026

Stealth Launch — Telegram First

Completed Instead of a big marketing splash, we quietly opened the doors through our Telegram bot. No ads, no press, no announcements — just the product, real users, and word-of-mouth. Every feature was battle-tested by a real community before the website ever went public.

  • Full platform launched inside Telegram (@AiNudeGirlsBot)
  • 25 bot languages from day one
  • Zero marketing spend — 100% word-of-mouth growth
  • Real-user feedback loop straight into development
  • Generation, model creation and payments all live in-bot
  • Community-driven iteration: ship, listen, improve, repeat

Creator Intelligence Platform (v4.1)

Completed AI-powered agency replacement with 9 intelligent features: fan segmentation, churn prediction, AI DM auto-responder, voice calls, companion mode, dynamic pricing, revenue forecasting, content prediction, and a unified intelligence dashboard.

  • Fan Segmentation & CRM — 5-factor scoring, LTV tracking, segment distribution
  • Churn Prediction — 5-factor risk scoring, auto win-back campaigns with DMs + promo codes
  • AI DM Auto-Responder — 3 modes (suggest/auto/hybrid), message classification, upsell engine
  • AI Voice Calls — Real-time Kokoro-82M TTS + Faster-Whisper STT, credits per minute, transcripts
  • AI Companion Mode — Persistent memory, relationship levels 1-10, in-chat image generation
  • PPV Dynamic Pricing — Per-fan pricing based on segment + purchase history
  • Revenue Forecasting — 3-month projections per stream with growth opportunities
  • Content Predictor — Engagement prediction + smart posting cadence analysis
  • Intelligence Dashboard — Unified view of all v4.1 features in one screen
  • 50 requirements, 10 phases, 4 database migrations, 20+ new backend files

Q2 2026: Organic Growth & Iteration

April — August 2026

3,000+ Users — 100% Organic

Completed The stealth community grew itself. No ads, no influencer deals, no paid promotion — just people telling people. Every one of the 3,000+ users found us through word-of-mouth, and their feedback shaped the platform daily.

  • 3,000+ registered users during stealth — zero ad spend
  • Pure word-of-mouth and community-driven growth
  • Daily feedback loop: user requests shipped within days
  • Referral program launched to reward the community
  • Retention and usage data guided every product decision

Smarter Generation Engine

Completed A complete overhaul of the generation brain: an AI language model now composes every prompt, backed by a hand-built knowledge base covering all 270+ categories, with visual quality control on the output.

  • LLM-powered prompt synthesis for every generation
  • Hand-curated knowledge base covering all 270+ categories
  • Automatic variety: same request, fresh look every time
  • Visual quality-control loop on rendered results
  • Sharper category accuracy across the whole library

Simple Flat-Rate Pricing

Completed Pricing rebuilt around fairness and predictability: one flat rate per image, identical on web and Telegram, with bundles of 1, 5, or 10 images.

  • One flat price per image — no hidden tiers
  • 1 / 5 / 10 image bundles
  • Web and Telegram pricing fully identical
  • 13 languages updated with the new pricing

Referral & Rewards Program

Completed Growth powered by the community — unique referral codes, automated credit rewards, and social login to make joining frictionless.

  • Unique referral code generation per user
  • Automated credit distribution on successful referrals
  • Referral tracking through OAuth sign-up flows
  • Social login on registration

LAUNCH: Website Goes Public

August 19, 2026

AINudeGirls.com Opens Worldwide

Launch Day After 22 months of development and a 4-month stealth run with 3,000+ users, the preview gate comes off and the full platform opens to the world — 270+ categories, custom model training in Beta, and 13 languages.

  • Public website launch — preview gate removed
  • NSFW Generator across 270+ categories
  • AI Video generation (Wan 2.2)
  • Create your own AI model (Beta)
  • 200+ AI model profiles with unique personalities
  • Full social platform: feeds, comments, likes, follows, notifications
  • Anonymous payment options: crypto, cards, wallets
  • Creator economy: dashboard, earnings, payouts, subscriptions
  • 13 languages across the whole platform
  • 3-stage QC pipeline ensuring image quality
  • Free credits for new users — no account required
  • 40+ pages, 90+ API endpoints, battle-tested in stealth

Q3 2026: Custom Models Beta

July — September 2026

Create Your Own AI Model — Beta Live

In Progress Custom AI model training is live in Beta right now. Describe your character and the platform trains a personal model with a consistent face you can use across every category.

  • Live in Beta today — on web (/create-model) and Telegram (/create)
  • Or let AI generate the full reference set from a description
  • Consistent face across all 270+ categories
  • Real-time training progress tracking
  • Private model gallery with access control

Q4 2026: The Comeback Season & Beyond

October — December 2026

Returning: AI Chat & Academy

Planned Two fan favourites are temporarily offline while we rebuild them from the ground up. Each one relaunches when it meets the quality bar the rest of the platform now sets.

  • AI Chat — smarter companions with deeper memory and personality
  • Academy — refreshed courses matched to the new tools
  • Each feature returns as soon as its rework is done — follow this page

AI Video Generation v2

Planned Next-generation video creation with extended length, text-to-video, style transfer, and batch processing capabilities.

  • Extended video length up to 30 seconds
  • Text-to-video generation (prompt only, no image needed)
  • Video style transfer and visual customization
  • Batch video processing for bulk creation
  • Video-to-video transformation pipeline
  • Higher resolution output (1080p target)

AiNG Sex Shop

Planned A curated adult shop built into the platform — toys, lingerie and accessories matched to your favourite categories and AI companions.

  • Curated selection of toys, lingerie & accessories
  • Recommendations matched to your favourite categories
  • Bundle deals for premium members
  • Discreet shipping & privacy-first checkout
  • Shop picks featured by your favourite AI models

Enhanced Creator Analytics

Planned Deep analytics for creators — content performance insights, audience demographics, optimal posting times, and revenue forecasting.

  • Advanced analytics dashboard with interactive charts
  • Content performance metrics (views, engagement, conversion)
  • Audience demographic breakdown
  • Optimal posting time suggestions
  • Revenue forecasting and trend analysis
  • A/B testing for content variants

Q1 2027: Mobile & Personalization

January — March 2027

Mobile App (iOS & Android)

Planned Native mobile experience with push notifications, optimized generation UI, offline content viewing, and biometric authentication.

  • Native iOS and Android applications
  • Push notifications for social interactions and content updates
  • Mobile-optimized AI generation interface
  • Offline content viewing (downloaded favorites)
  • Biometric authentication (Face ID, Touch ID, fingerprint)
  • Swipe-optimized discovery interface

Advanced Personalization Engine

Planned Machine learning-powered content personalization — learning from user preferences, engagement patterns, and explicit taste settings.

  • ML-based content recommendation refinement
  • Implicit preference learning from engagement patterns
  • Explicit taste settings (preferred categories, body types, styles)
  • Personalized model suggestions based on interaction history
  • "Similar models" and "You might also like" features

Content Scheduling & Auto-posting

Planned Schedule posts in advance — calendar view, queue management, auto-generation of daily content, and timezone-aware publishing.

  • Visual content calendar with drag-and-drop
  • Post queue with priority ordering
  • Timezone-aware scheduling (per-user preference)
  • Auto-generation pipelines for daily fresh content
  • Recurring post templates

Community Features v2

Planned Deeper community engagement — user galleries, challenges, competitions, creator collaboration, and social sharing tools.

  • User-generated content galleries (opt-in sharing)
  • Weekly community challenges with prizes
  • Creator collaboration tools (shared content, cross-promotion)
  • Advanced social sharing with preview cards
  • Community voting and curation system
  • Badges and achievement system

Q2 2027: Platform & Scale

April — June 2027

Public Developer API

Planned RESTful API for third-party integrations — full documentation, SDKs for major languages, webhook support, batch processing, and a free tier for developers.

  • RESTful API with comprehensive OpenAPI documentation
  • SDKs for Python, JavaScript, and Go
  • Webhook integrations for event-driven workflows
  • Batch processing endpoints for bulk generation
  • Rate-limited free tier for developers
  • API key management and usage analytics

Real-Time Features

Planned Live interaction capabilities — real-time content drops, live generation sessions, and instant notifications across all devices.

  • Live AI generation sessions (watch creation in real-time)
  • Real-time content drops with countdown timers
  • Cross-device notification sync
  • Live chat rooms for community events
  • Instant content delivery via WebSocket

Multi-Model AI Engine

Planned Support for multiple AI model architectures — SDXL, Flux, and emerging diffusion models alongside existing SD 1.5 infrastructure.

  • SDXL model support for higher resolution output
  • Flux model integration for next-gen quality
  • Model routing: auto-select best model per task
  • A/B quality comparison between model architectures
  • Cloud GPU scaling for heavy workloads

Cross-Post to Other Platforms

Planned Publish your content once and share it to OnlyFans, Fansly, Twitter/X, Reddit and 20+ other sites from a single place.

  • One-click publish to 25+ platforms
  • Per-platform captions & scheduling
  • Securely connect creator & social accounts

Direct Messages

Planned Private one-to-one messaging between fans and creators, with pay-per-view unlocks and real-time delivery.

  • Real-time DM inbox
  • Pay-per-view media unlocks
  • Unread badges & notifications

AI Voice Calls

Planned Real-time voice conversations with AI companions, in your model's own cloned voice.

  • Natural real-time voice chat
  • Per-model cloned voices
  • Call history & credits

Building in Public

We believe in complete transparency. This roadmap documents every major milestone across 22 months of development — from our first prototype, through the Telegram stealth launch and 3,000+ organically grown users, to worldwide release. It's updated regularly as we ship features and define the future.