Audio Transcription API

A Go (Golang) backend service for transcribing audio files using FFmpeg for processing and the Groq API for transcription.

Features

• Upload and transcribe audio files (MP3, WAV, FLAC, M4A, etc.)
• Audio preprocessing with FFmpeg for optimal transcription quality
• Audio chunking for large files to improve transcription accuracy
• Parallel processing of audio chunks for faster results
• RESTful API for easy integration with frontend applications

Tech Stack

• Go (Golang)
• Gin framework for HTTP routing and middleware
• FFmpeg for audio processing
• Groq API for transcription (using the Whisper model)
• UUID generation for temporary file management

Requirements

• Go 1.16+
• FFmpeg installed on the system
• FFprobe installed on the system (comes with FFmpeg)
• Groq API key

Installation

1. Clone the repository:

   git clone <repository-url>
   cd <repository-directory>

2. Install dependencies:

   go mod download

3. Add your Groq API key to the transcribeAudio function in main.go:

   // Replace the empty string with your API key
   apiKey := "your_groq_api_key_here"

Running the Server

Start the server:

go run main.go

The server will run on port 8080 by default.

API Endpoints

Transcribe Audio

Endpoint: POST /api/transcribe

Request:

• Content-Type: multipart/form-data
• Body:
  • file: Audio file (MP3, WAV, FLAC, M4A, etc.)

Response:

{
  "transcription": "This is the transcribed text from the audio file..."
}

Error Response:

{
  "error": "Error message describing what went wrong"
}

Implementation Details

Audio Processing Pipeline

1. File Upload: The API accepts audio file uploads via a multipart form.
2. Preprocessing: The uploaded audio is preprocessed using FFmpeg:
   • Converted to 16kHz sample rate
   • Reduced to mono channel
   • Converted to FLAC format for optimal transcription
3. Chunking: Large audio files are split into manageable chunks (2 minutes each with a 1-second overlap)
4. Parallel Processing: Multiple chunks are transcribed simultaneously (limited to 5 concurrent operations)
5. Transcription: Each chunk is sent to the Groq API for transcription using the distil-whisper-large-v3-en model
6. Combination: Results are combined and returned as a complete transcription

Code Structure

• Main Function: Sets up the Gin router with CORS configuration and defines the API routes
• transcribeAudio: The main handler function that orchestrates the audio processing and transcription
• preprocessAudioFile: Processes audio files to prepare them for transcription
• getAudioChunkData: Analyzes audio files to determine chunking parameters
• chunkifyAudioFile: Splits large audio files into smaller chunks
• createAudioChunkFile: Creates individual audio chunk files
• transcribeChunk: Sends audio chunks to the Groq API for transcription
• deleteFiles: Cleans up temporary files

Extending the API

To add additional functionality:

1. Add new route handlers in main.go
2. Create helper functions for new features
3. Update the error handling as needed

CORS Configuration

The API is configured to accept requests from:

• http://localhost:5173 (default Vue.js development server)

If you need to allow additional origins, update the CORS configuration in the main function.

Error Handling

The API includes comprehensive error handling:

• Validation errors for missing files or bad requests
• Audio processing errors from FFmpeg operations
• Transcription errors from the Groq API
• Cleanup of temporary files even in error cases

Performance Considerations

• Concurrency Control: The API limits the number of concurrent transcription operations to 5 to prevent overloading the system or hitting API rate limits.
• CPU Utilization: Audio chunk processing uses the available CPU cores (with a default of 4 if GOMAXPROCS is not set)
• Temporary File Management: All temporary files are properly cleaned up after processing.

License

MIT License