peren55

All‑in‑one AI Studio that creates apps, games, WhatsApp bots, viral avatars and video content in seconds with full Hebrew & English support.

TL;DR (for recruiters)

• Full‑stack monorepo (React frontend + Node/Express backend)
• AI generation pipelines for apps/games/media
• Realtime via Socket.io
• Viral avatars: TTS + (optional) local Wav2Lip lip‑sync service
• Production configs included (Railway/Render/Netlify/Vercel)

Demo

• MP4: https://raw.githubusercontent.com/oren40/peren55/main/public/assets/peren55.MP4

What this repo contains

This is a monorepo:

• Frontend (root): React 18 + CRACO + TypeScript UI
• Backend (peren55-backend/): Node.js + Express API (Prisma, Socket.io, AI orchestration)
• Local AI services (peren55-backend/ai/): Python FastAPI services (Talking Head / Wav2Lip) + vendor models

Key features

• AI app/game generation
  • Multi-agent generation pipeline (peren55-backend/src/agents, peren55-backend/src/ai)
  • Templates + enrichment + validations
• Viral Avatars (Talking Head / Lip Sync)
  • TTS (Edge‑TTS / fallback)
  • Local Wav2Lip FastAPI service for lip-sync video generation
• Media tools
  • Image/video tooling and export workflows
• Auth & payments
  • OAuth (Google/GitHub)
  • JWT sessions
  • PayPal / Stripe integration
• Realtime
  • Socket.io events + collaboration flows

Multi-agent pipeline (high level)

• Intent + language routing: quick HE/EN detection + fallback to LLM intent detection (peren55-backend/src/services/builderOrchestrator.ts).
• Spec generation: convert prompt -> structured app spec (spec_generation).
• IR + scaffolding: generate intermediate representation + scaffold a runnable project (ir_generation, scaffold).
• Agent specialization: category agents (App/Game/Media/Tool/etc.) live in peren55-backend/src/agents/.
• Realtime progress: every step emits progress events (job state + % + step) over Socket.io to keep the UI responsive.

Tech stack

Frontend

• React 18, TypeScript
• CRACO (Create React App customization)
• TailwindCSS
• MUI, Mantine
• Three.js (@react-three/fiber, @react-three/drei) for 3D
• i18next for Hebrew/English

Backend

• Node.js 18+, Express
• Prisma (Postgres)
• Socket.io
• Cloudinary (media hosting)
• FFmpeg (media processing)

Architecture overview

graph TD
  A[Frontend React TypeScript] -->|HTTP API| B[Backend Node Express]
  A -->|Realtime| S[Socketio]
  B -->|Orchestration| O[Builder Orchestrator Agents]
  O --> L[LLM Providers]
  B --> DB[Postgres Prisma]
  B --> M[Media Cloudinary FFmpeg]
  B --> W[Optional FastAPI Wav2Lip]
  S --> B

Loading

Local AI (optional, for self-hosted viral avatars)

• Python 3.10+
• FastAPI + Uvicorn
• Wav2Lip (vendorized under peren55-backend/ai/talking_head/vendor/Wav2Lip)
• edge-tts

Local development (Windows)

Prerequisites

• Node.js 18+
• Python 3.10+ (only if you want local Wav2Lip / Talking Head)
• FFmpeg (recommended)

1) Install

npm install
npm --prefix ./peren55-backend install

2) Configure environment

Create these files:

• ./.env from ./.env.example
• ./peren55-backend/.env from ./peren55-backend/.env.example

3) Run everything

This starts:

• Frontend on http://localhost:3001
• Backend on http://localhost:3002
• Wav2Lip FastAPI on http://127.0.0.1:8001
• Prisma Studio on http://localhost:5555

npm run dev:windows

Useful scripts

• npm run dev:windows – start frontend + backend + wav2lip + prisma studio
• npm run dev:frontend:windows – start frontend only
• npm run server:windows – start backend only
• npm --prefix ./peren55-backend run db:studio – Prisma Studio

Useful URLs

• Frontend: http://localhost:3001
• Backend health: http://localhost:3002/api/health
• Backend docs: http://localhost:3002/api-docs
• Wav2Lip health: http://127.0.0.1:8001/api/health
• Prisma Studio: http://localhost:5555

Project structure (high level)

.
├─ src/                       # Frontend React app
├─ public/                    # Static assets
├─ peren55-backend/
│  ├─ src/                    # Express API + services
│  ├─ prisma/                 # Prisma schema/migrations
│  ├─ ai/
│  │  └─ talking_head/        # FastAPI (Wav2Lip) service
│  └─ temp/                   # Runtime temp files (ignored)
└─ render.yaml / railway.toml  # Deployment configs

Notable modules

• Backend entrypoint: peren55-backend/src/server.js
• Self-hosted viral video pipeline: peren55-backend/src/routes/selfHostedVideo.js
• Local Wav2Lip service (FastAPI): peren55-backend/ai/talking_head/main.py
• Wav2Lip processing implementation: peren55-backend/ai/talking_head/service_wav2lip.py
• Frontend hub: src/components/MediaToolsHub.tsx

Key API endpoints

• GET /api/health – backend health check
• POST /api/self-hosted/create-audio – TTS generation
• POST /api/self-hosted/create-video – avatar video creation via Wav2Lip
• POST /api/self-hosted/generate-viral-video – full pipeline (Text -> TTS -> LipSync)

Deployment

This repo contains configs for multiple providers:

• Railway: railway.toml + nixpacks.toml
• Render: render.yaml + render-build.sh
• Netlify: netlify.toml (frontend)
• Vercel: vercel.json (frontend)

Recommended approach:

• Frontend: Netlify/Vercel (static)
• Backend: Railway/Render (Node)
• Viral avatars (local Wav2Lip): run on the same machine/instance as backend when using the local FastAPI service

Challenges & solutions

• Realtime build progress: long-running AI jobs need continuous UI feedback.
  • Solution: job orchestration emits structured progress events (step + cumulative progress) via Socket.io so the frontend can show status in real time.
• Mixed runtimes (Node + Python): keeping media pipelines stable across environments.
  • Solution: isolate optional Wav2Lip/Talking-Head behind a FastAPI service boundary and treat it as an external dependency with healthchecks.

Security

• Never commit .env files.
• Use .env.example files and configure real secrets in your hosting provider’s environment variables.

CI/CD

GitHub Actions workflows live in .github/workflows/.

License

MIT (see LICENSE).

Contact

https://www.linkedin.com/in/oren-peretz

oren.peretz40@gmail.com