readme

README.md

@@ -0,0 +1,41 @@
+---
+title: ATC TTS MOS Evaluation
+emoji: 🎙️
+colorFrom: blue
+colorTo: purple
+sdk: gradio
+sdk_version: "4.44.0"
+app_file: frontend/app.py
+pinned: false
+---
+
+# ATC TTS MOS Evaluation
+
+An interactive evaluation interface for rating Air Traffic Control (ATC) Text-to-Speech (TTS) models using Mean Opinion Score (MOS) methodology.
+
+## Features
+
+- **Sample Audio**: Listen to reference clips from different TTS models
+- **MOS Rating**: Rate individual clips on clarity, pronunciation, prosody, naturalness, and overall quality
+- **A/B Model Comparison**: Compare different TTS models side-by-side
+- **A/B Gender Comparison**: Compare male vs female voices from the same model
+- **Data Collection**: Results are automatically saved to Hugging Face datasets
+
+## How to Use
+
+1. **Introduction**: Read the evaluation guidelines
+2. **Sample Audio**: Listen to reference clips to familiarize yourself with the audio quality
+3. **MOS Rating**: Rate 18 clips across different models and speakers
+4. **Model Comparison**: Choose between different TTS models for the same content
+5. **Gender Comparison**: Compare male and female voices
+6. **Submit**: Complete the evaluation and submit your responses
+
+## Models Evaluated
+
+- **CSM**: Custom Speech Model
+- **XTTS**: XTTSv2 Model  
+- **Orpheus**: Orpheus TTS Model
+
+## Data Storage
+
+All evaluation responses are stored in the `aether-raid/atc-tts-mos-ratings` dataset for research purposes.