AsyncChunkPy: Near-Realtime Python Speech-to-Text App

AsyncChunkPy is a Python application that provides near-realtime speech-to-text transcription using chunked audio processing and asynchronous transcription. It leverages the power of AssemblyAI's async transcription API to deliver high-quality transcriptions at near real-time speeds.

Features

• Real-time audio recording and chunking
• Voice Activity Detection (VAD) for intelligent chunk processing
• Asynchronous transcription using AssemblyAI API
• Ordered transcript logging
• Configurable chunk size and silence threshold
• Support for multiple languages

Key Benefits

• Access to AssemblyAI's powerful Universal-2 model for English and Universal-1 model for Spanish and German
• Support for all non-English languages available in AssemblyAI's async transcription service
• Higher accuracy compared to real-time transcription models
• More cost-effective than real-time transcription services
• Near real-time performance with the quality of async transcription

Prerequisites

• Python 3.7 or later
• pip (Python package installer)
• AssemblyAI API key (You can # for an AssemblyAI account and get your API key from your dashboard.)

Installation

1. Clone the repository:

   git clone https://github.com/AssemblyAI-Solutions/async-chunk-py.git
   cd async-chunk-py
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Create a .env file in the root directory and add your AssemblyAI API key:

   ASSEMBLYAI_API_KEY=your_api_key_here
   

Usage

1. Start the application:

   python main.py
   

2. Speak into your microphone. The application will record and transcribe your speech in near-realtime.

3. Press Ctrl+C to stop the recording and see the final transcript.

Configuration

You can modify the following parameters in config.py:

• CHUNK_SIZE: Size of each audio chunk in bytes
• CHUNK_DURATION_MS: Duration of each audio chunk in milliseconds (default: 5000ms)
• SILENCE_THRESHOLD_MS: Duration of silence required to trigger chunk processing (default: 600ms)

To change the language or enable language detection, modify the transcription_worker.py file:

• Set language_code='en' to the desired language code in the transcribe method, or
• Add language_detection=True to enable automatic language detection

Voice Activity Detection (VAD)

This project uses the py-webrtcvad library for VAD. You can adjust VAD parameters by modifying the Vad configuration in audio_recorder.py. For more information on VAD parameters, visit the py-webrtcvad GitHub repository.

Project Structure

• main.py: Main application file handling coordination between audio recording and transcription.
• audio_recorder.py: Handles audio recording and Voice Activity Detection.
• transcription_worker.py: Worker for handling transcription tasks using AssemblyAI API.
• config.py: Configuration file for various parameters.

Acknowledgments

• py-webrtcvad for the Voice Activity Detection functionality

Troubleshooting

If you encounter any issues with audio recording or transcription, ensure that:

1. Your microphone is properly connected and selected as the input device.
2. Your AssemblyAI API key is correctly set in the .env file.
3. You have a stable internet connection for API communication.

For any other issues, please check the console output for error messages and refer to the documentation of the individual dependencies if needed.