0xkhingx/Moodmix

Capstone project for Electric Sheep Africa --- title: MoodMix emoji: 🎵 colorFrom: gray colorTo: green sdk: docker app_port: 7860 --- # MoodMix A mood-to-Spotify playlist generator that uses machine learning to detect emotion from speech or text input and creates a curated Spotify playlist based on the detected mood. ## Features - **Speech Emotion Recognition** — Record your voice via microphone; a CNN analyzes mel-spectrogram features to classify emotion (96.4% accuracy) - **Text Emotion Classification** — Type how you're feeling; a TF-IDF + Logistic Regression model classifies the sentiment (86.7% accuracy) - **Spotify Integration** — Automatically creates a playlist with 10 tracks matching the detected mood using Spotify's search API - **Streamlit UI** — Dark-themed interface with Font Awesome icons, mood-colored cards, and track previews ## Emotion Classes | Emotion | Search Query | Description | |---------|-------------|-------------| | Happy | upbeat, feel-good | Upbeat and joyful | | Sad | emotional, melancholy | Melancholic and reflective | | Angry | intense, heavy | Intense and powerful | | Calm | relaxing, peaceful | Peaceful and relaxed | | Fearful | dark, tense, suspenseful | Tense and uneasy | | Surprised | surprising, energetic | Bright and unexpected | | Disgusted | dark, industrial | Harsh and abrasive | ## Project Structure ``` moodmix/ ├── app.py # Streamlit entry point ├── requirements.txt # Dependencies ├── .env # Spotify API credentials (not tracked) ├── .gitignore ├── config/ │ └── mood_features.json # Mood-to-query mappings and icons ├── data/ │ ├── TESS/ # TESS audio dataset (not tracked, ~2800 WAVs) │ ├── text_emotion/ # Text emotion dataset (train/test/val) │ └── processed/ # Preprocessed numpy arrays (not tracked) ├── models/ │ ├── speech_mood_model.pth # Trained CNN weights (32 MB) │ ├── text_mood_pipeline.pkl# Text model + vectorizer │ └── text_history.pkl # Text train …