Discover music through AI.
Generate highly relevant song recommendations, refine them conversationally, and instantly turn them into a Spotify playlist.
-
🎵 AI-powered recommendations
- Input any song (e.g. "The Weeknd - Blinding Lights")
- Get 20 highly relevant tracks
-
🧠 Conversational refinement
- Refine results naturally:
- "darker"
- "more synthwave"
- "less mainstream"
- "female vocals"
- Session-aware context (remembers previous state)
- Refine results naturally:
-
🔗 YouTube integration
- One-click search for every recommended song
-
🎼 Spotify playlist creation
- Instantly create a playlist with all recommended tracks
- Automatic matching via Spotify API
-
🎨 Modern UI
- Clean dark design
- AI loading animation with typing effect
- Smooth UX and transitions
- User inputs a song
- Gemini AI:
- Detects song & artist
- Understands vibe, mood, production
- Generates 20 similar tracks + playlist name
- User can refine results conversationally
- App optionally creates a Spotify playlist
Frontend (React + Vite) ↓ Backend (Node.js + Express) ↓ Gemini API (AI recommendations) ↓ Spotify API (playlist creation)
- React
- Vite
- CSS (custom + Tailwind-inspired styling)
- Node.js
- Express
- Zod (validation)
- Google Gemini API
- Spotify Web API
music-ai-gemini-only/ ├── src/ # Backend │ ├── services/ │ ├── routes/ │ ├── session/ │ ├── utils/ │ └── index.js │ ├── vibe-ai-ui/ # Frontend │ ├── src/ │ ├── index.css │ └── App.jsx │ ├── .env └── README.md
git clone https://github.com/your-username/vibe-ai.git
cd vibe-ainpm install
npm run servercd vibe-ai-ui
npm install
npm run devCreate .env in root:
GEMINI_API_KEY=your_key_here
SPOTIFY_CLIENT_ID=your_id
SPOTIFY_CLIENT_SECRET=your_secret
SPOTIFY_REDIRECT_URI=http://127.0.0.1:8888/callback{
"query": "The Weeknd - Blinding Lights"
}{
"refinement": "darker, more synthwave"
}Creates Spotify playlist from current session.
- AI-first product design (not just API wrapper)
- Session-based conversational refinement
- Real-world integrations (Spotify)
- Clean UX with loading states and feedback
- Fully functional end-to-end system
- Real YouTube video embedding (not just search)
- User accounts & saved playlists
- Deploy (Vercel + Render)
- Mobile app (React Native)
- Better recommendation ranking (hybrid AI + API)
Built by Konstantinos Tsaousidis
MIT
