Voxion

Voxion is a powerful speech recognition and annotation system that supports multilingual speech recognition and annotation management. It uses the OpenAI Whisper model for speech recognition and integrates with the Codatta platform for data annotation management.

Features

• Multilingual speech recognition support (Chinese, English, Japanese, Korean, etc.)
• Real-time voice recording and recognition
• Multiple Whisper models available (from fast to high accuracy)
• Seamless integration with Codatta platform
• Offline model download and usage support
• User-friendly web interface

Quick Start

1. Clone the repository:

git clone https://github.com/paulhandle/voxion.git
cd voxion

2. Create and activate virtual environment:

python -m venv venv
source venv/bin/activate  # Linux/Mac
# or
venv\Scripts\activate  # Windows

3. Install dependencies:

pip install -r requirements.txt

4. Start the application:

python -m flask run --port=5002 --debug

5. Access the application:

http://localhost:5002

Technology Stack

• Backend: Python Flask
• Speech Recognition: OpenAI Whisper
• Frontend: HTML5, JavaScript
• Audio Processing: MediaRecorder API

Codatta Integration APIs

1. Redirect from Codatta to ASR

Handles user redirection from Codatta platform to ASR system for annotation.

Endpoint

GET /asr/task

Request Parameters

Parameter	Type	Required	Description
token	string	Yes	Codatta user authentication token
task_id	string	Yes	Codatta task ID

Response Status Codes

• 200: Success, redirects to ASR annotation page
• 401: Invalid token
• 404: Task not found
• 400: Missing required parameters

Example

http://localhost:5002/asr/task?token=user_token_123&task_id=task_456

2. Submit Annotation to Codatta

Submits annotation data back to Codatta platform.

Endpoint

POST https://api.codatta.com/v1/annotations

Request Headers

Authorization: Bearer <token>
Content-Type: application/json

Request Body

{
    "task_id": "string",      // Codatta task ID
    "audio_data": "string",   // Base64 encoded audio data
    "transcription": "string", // Transcribed text
    "language": "string",     // Audio language
    "model": "string",        // Model used
    "timestamp": "string"     // ISO format timestamp
}

Response Status Codes

• 200: Successfully submitted
• 401: Invalid token
• 404: Task not found
• 400: Invalid request data format

Response Example

{
    "status": "success",
    "annotation_id": "annotation_789"
}

Development Mode

For testing Codatta integration in development environment:

1. Enable mock mode:

export MOCK_CODATTA_API=true

2. Use test token (prefixed with "mock_"):

http://localhost:5002/asr/task?token=mock_user_123&task_id=task_456

3. Mock responses are configured in config.py:

MOCK_RESPONSES = {
    'submit_annotation': {
        'status': 'success',
        'annotation_id': 'mock_annotation_123'
    }
}

Production Configuration

When deploying to production:

1. Set correct API URL:

export CODATTA_API_BASE_URL=https://api.codatta.com/v1

2. Disable mock mode:

export MOCK_CODATTA_API=false

3. Set secure session key:

export FLASK_SECRET_KEY=your_secure_key

License

This project is licensed under the MIT License - see the LICENSE file for details.