🎤 Khmer Whisper ASR - G5 Project

Automatic Speech Recognition for the Khmer language using OpenAI's Whisper model, fine-tuned on Khmer audio data.

Features

✨ Real-time Transcription - Upload audio files or record directly from microphone 🇰🇭 Khmer Optimized - Fine-tuned on authentic Khmer speech data ⚡ Fast Processing - Efficient Whisper Small model 🎯 High Accuracy - Specialized training on Khmer phonetics

How to Use

Upload Audio: Click "Upload" to select an audio file (WAV, MP3, OGG, etc.)
Record Audio: Click the microphone icon to record directly
Get Transcription: The model will transcribe your audio to Khmer text

Model Details

Base Model: OpenAI Whisper Small
Language: Khmer (km)
Fine-tuning: Trained on Khmer voice data
Framework: Hugging Face Transformers

About the G5 Project

This is part of the G5 Khmer ASR project aimed at improving automatic speech recognition for the Khmer language.

Technical Stack

🤗 Hugging Face Transformers
🎵 Librosa (audio processing)
🐍 Python 3.8+
🚀 Gradio (interface)