Local-first AI creative pipeline for persona-based image and video generation with identity preservation, prompt orchestration, and quality scoring.
Every image and video below was generated 100% by this pipeline. Same persona, consistent identity across all outputs.
👆 8 images, 1 persona, different styles — the face stays consistent because of our identity lock engine.
Image → 8s video · VEO 3.1 |
Urban lifestyle |
Fashion editorial |
The pipeline can generate multi-shot sequences — multiple cinematic clips from different angles, seamlessly chained:
Close-up hook shot |
Mid-range portrait |
Movement tracking |
⭐ If this is useful to you, give it a star! It helps others discover it.
Most AI image generators give you a prompt box and a button. Persona Media Studio is a full content pipeline — from prompt planning to quality scoring.
- ✅ Image + Video generation — Gemini for images, VEO 3.1 for cinematic 9:16 video
- ✅ Identity preservation — same face across all outputs using reference photos
- ✅ 7 visual engines — lighting, shadow, motion, weather, outfit, scene validation, visual context
- ✅ Smart content planner — weighted pillar selection, anti-repeat scene picker, caption generator
- ✅ AI quality scoring — 5-channel analysis (face, eyes, lighting, motion, composition)
- ✅ Autopilot mode — batch generate with automatic planning
- ✅ Publish guard — quality gate blocks low-quality outputs
- ✅ Your own API key — pay Google directly, no middleman
- This tool is intended for user-owned or consented creative assets only.
- It does not include impersonation, scraping, or automated publishing features.
- API keys stay user-controlled (bring-your-own-key model).
- Generated assets are stored locally by default.
- Do not use this tool to generate non-consensual content of real people.
Photorealistic images with identity-locked reference photos. Upload one photo → get consistent identity across hundreds of generations.
8-second cinematic 9:16 videos from any generated image using Google VEO 3.1. Perfect for Reels, TikTok, Shorts.
Ported from a production autonomous agent pipeline:
- Lighting Engine — golden hour, studio, neon, based on location & time
- Shadow Engine — directional shadows matching the lighting
- Motion Engine — motion profiles (subtle, dynamic, static)
- Weather Engine — weather validation & visual effects
- Outfit Bias Engine — outfit adjustments for scene/weather
- Scene Validator — validates shot/scene combinations
- Visual Context Orchestrator — combines all engines into one cinematic prompt
- Weighted pillar selection — balances content types (lifestyle, urban, music...)
- Anti-repeat scene picker — never generates the same location/outfit/time twice in a row
- Shot picker — weighted archetype selection with constraints
- Caption generator — pillar-aware captions with soft CTAs
Python FastAPI microservice that analyzes generated videos across 5 channels:
- Face stability · Eye engagement · Lighting consistency · Motion smoothness · Composition
Quality gate with persona-specific thresholds — blocks low-quality content from publishing.
Batch generate content automatically. Pick a persona, set the count, hit go. Planner decides style, location, shot — you just collect the results.
git clone https://github.com/office233/MoneyPrinterPro.git
cd MoneyPrinterPro
npm install- Go to Google AI Studio
- Click "Create API Key"
- Copy the key
npm run dev- Open
http://localhost:3000 - Go to ⚙️ Settings → paste your Gemini API key → Test Connection
- Create a persona → upload a reference photo
- Generate images or videos 🎉
cd scoring-service
pip install -r requirements.txt
uvicorn main:app --port 8000You pay Google directly. No middleman. No markup. No subscription.
| What | Model | Cost |
|---|---|---|
| Image (512px) | Gemini Flash Image | ~$0.045 |
| Image (1024px) | Gemini Flash Image | ~$0.067 |
| Video (8s 9:16) | VEO 3.1 | ~$0.50 |
$10 ≈ 150 images or 20 videos. Your data stays on your machine.
money-printer-pro/
├── src/app/ # Next.js pages
│ ├── page.jsx # Dashboard (job history + stats)
│ ├── generate/ # Generation page (image/video/both)
│ ├── autopilot/ # Autopilot batch generation
│ ├── settings/ # API key settings
│ ├── personas/new/ # Create persona
│ └── api/ # Backend routes
│ ├── generate/ # Image generation
│ ├── generate-video/ # Video generation (VEO)
│ ├── jobs/ # Job history
│ ├── autopilot/ # Autopilot controller
│ └── score/ # Quality scoring
├── src/lib/
│ ├── engines/ # 7 visual engines
│ ├── planner/ # Content planner
│ ├── video/ # VEO video module
│ ├── scoring/ # Scoring client
│ ├── db.js # SQLite database
│ └── publish-guard.js # Quality gate
├── src/config/ # Engine config JSONs
├── scoring-service/ # Python FastAPI scoring
├── examples/ # Generated examples (in repo)
├── assets/ # Reference & generated (gitignored)
└── data/ # SQLite database (gitignored)
| Layer | Technology |
|---|---|
| Frontend | Next.js 16, React 19 |
| AI — Images | Google Gemini API (@google/genai) |
| AI — Video | Google VEO 3.1 (Vertex AI) |
| Engines | 7 visual context engines (pure JS) |
| Database | SQLite via better-sqlite3 |
| Scoring | Python FastAPI + OpenCV + InsightFace |
| Styling | Vanilla CSS (dark Linear-inspired theme) |
- API keys live only in your browser's
localStorage. Sent viax-api-keyheader only. - No server-side fallback key. No key = HTTP 401.
- All inputs validated server-side — path traversal protection, MIME checking, body size limits.
- No telemetry, no tracking, no analytics.
- Everything stays on your machine.
- The quality scoring service is experimental and requires manual setup.
- Rate limiting is local/in-memory and not suitable for distributed deployment.
- Production authentication and team-management features are not implemented.
- Test coverage should be expanded before production use.
- The video scoring pipeline requires additional Python dependencies.
MIT — Use it however you want.
⭐ Star this repo if you find it useful!