🎵 Music-Inspired Image Stylization App

Transform your photos with AI-powered visual elements that match the mood and vibe of your favourite Spotify songs.

About:
A web app that analyses Spotify songs and uses AI to stylise your images, blending music and visual art in a unique, interactive way.

Live Demo

---

📝 Project Overview

This app lets users upload an image and select a Spotify song. Using AI, it analyses the song's mood and transforms the image to visually match the song's emotional characteristics. The result: a personalised, music-inspired artwork.

⚙️ How It Works

1. Image Upload: User uploads a photo.
2. Song Search: User searches and selects a song from Spotify.
3. Song Analysis: The backend fetches the song's mood, keywords, and vibe using Spotify's audio features and AI-powered analysis.
4. Mood Mapping: These features are mapped to visual style prompts using a combination of AI and custom logic.
5. AI Generation: Stable Diffusion (via Replicate) stylises the image based on the prompt.
6. Result Display: Both original and stylised images are shown to the user.

🏗️ Architecture & Tech Stack

Frontend

• Next.js 14 (React, TypeScript)
• TailwindCSS (Styling)
• React Hook Form (Form handling)
• Axios (HTTP requests)

Backend

• FastAPI (Python 3.8+)
• Pillow (Image processing)
• python-multipart (File uploads)
• Spotipy (Spotify API integration)
• Replicate API (Stable Diffusion & Llama-2 for AI analysis)
• Requests (HTTP requests)
• OpenAPI/Swagger (Auto-generated docs)

✨ Key Features

• Upload any photo (JPEG, PNG, WebP, up to 10MB)
• Search and select songs from Spotify
• AI analyses song mood and generates visual prompts
• Stable Diffusion transforms the image
• Responsive, modern UI
• Side-by-side comparison of original and stylised images
• Real-time feedback and error handling

🎨 Mood-to-Style Mapping (AI Logic)

• The backend uses an AI model (Llama-2 via Replicate) to analyse the selected song and extract its mood, keywords, and a description.
• A custom prompt engineering function then translates these results into concrete visual elements (e.g., posters, lighting, decorations) that match the song's vibe.
• Special logic ensures that only appropriate elements are suggested (e.g., neon lights only for electronic genres, blue lighting for sad songs, etc.).
• The final style prompt is sent to Stable Diffusion (via Replicate) to generate the stylised image.

🙋‍♂️ Built by Me — powered by Cursor