Audio Transcription Application

A simple and elegant web application for audio transcription using Whisper AI. Record audio directly from your microphone or upload WAV files to get instant transcriptions.

Features

• 🎙️ Audio Recording: Record audio directly from your browser
• 📁 File Upload: Upload existing WAV files
• 🌍 Language Selection: Choose the audio language for better transcription accuracy
• ✨ AI Transcription: Powered by OpenAI-compatible Whisper API
• 📝 AI Summarization: Get concise summaries of transcribed text using LLM
• 🎨 Modern UI: Clean, responsive design with great UX
• 🐳 Container-Ready: Easy deployment with Podman/Docker

Architecture

• Backend: Go server (standard library only, ~300 lines)
  • Serves static files
  • Acts as a gateway to the transcription API
  • Acts as a gateway to the LLM API for summarization
• Frontend: Pure JavaScript with PatternFly UI
  • Red Hat PatternFly design system for professional UI
  • Audio recording with MediaRecorder API
  • WAV file conversion and upload
  • Real-time transcription display
  • AI-powered summarization

Prerequisites

• Podman or Docker
• An OpenAI-compatible Whisper API endpoint (e.g., faster-whisper, whisper.cpp)
• An OpenAI-compatible LLM API endpoint (e.g., vLLM, Ollama, OpenAI API)

Quick Start

1. Set the environment variables

export AUDIO_INFERENCE_URL=http://your-whisper-server:8000
export AUDIO_MODEL_NAME=whisper-1  # Optional, defaults to whisper-1
export LLM_INFERENCE_URL=http://your-llm-server:8000
export LLM_MODEL_NAME=gpt-3.5-turbo  # Optional, defaults to gpt-3.5-turbo

2. Build the application

make build

3. Run the application

make run

The application will be available at http://localhost:8080

Makefile Commands

Command	Description
make build	Build the container image
make run	Run the application
make stop	Stop and remove the container
make clean	Stop container and remove image
make logs	Show application logs
make restart	Restart the application
make help	Show available commands

Configuration

Environment Variables

• AUDIO_INFERENCE_URL (required): URL of the OpenAI-compatible Whisper API server

  • Example: http://localhost:8000
  • The server expects the API to be available at /v1/audio/transcriptions

• AUDIO_MODEL_NAME (optional): Model name to use for transcription

  • Default: whisper-1
  • Example: whisper-large-v3, whisper-medium, etc.

• LLM_INFERENCE_URL (required): URL of the OpenAI-compatible LLM API server

  • Example: http://localhost:8001
  • The server expects the API to be available at /v1/chat/completions

• LLM_MODEL_NAME (optional): LLM model name to use for summarization

  • Default: gpt-3.5-turbo
  • Example: gpt-4, llama3, mistral, etc.

• PORT (optional): Port for the application (default: 8080)

Usage

Recording Audio

1. Click "Start Recording" to begin capturing audio
2. Click "Stop Recording" when finished
3. Review the recording using the playback controls
4. Optionally select the audio language from the dropdown (or leave as "Auto-detect")
5. Click "Transcribe Audio" to get the transcription

Uploading Files

1. Click "Choose WAV File"
2. Select a WAV file from your computer
3. Optionally select the audio language from the dropdown (or leave as "Auto-detect")
4. Click "Transcribe Audio" to get the transcription

Audio Language Selection (Optional Hint)

The language dropdown serves as an optional hint to improve transcription accuracy:

• Auto-detect (default): Whisper automatically detects and transcribes in the original language
• Specific language: Provide a hint about the audio's spoken language for better accuracy

Important:

• This parameter does NOT translate the audio
• The transcription will always be in the same language as the audio
• Whisper can only translate TO English (not implemented in this app)
• The language hint simply helps Whisper recognize the correct language faster and more accurately

Viewing Results

• The transcription appears in a text box (plain text format)
• Click "Summarize" to generate an AI summary of the transcription
• The summary displays with rich Markdown formatting:
  • Bold text, italic text
  • Headers and subheaders
  • Bullet lists and numbered lists
  • Inline code formatting
• Click "Copy to Clipboard" to copy the transcription text
• Click "Copy Summary" to copy the summary text (plain text)
• Click "New Transcription" to start over

API Endpoints

The Go server exposes the following endpoints:

• GET / - Main application page
• GET /static/* - Static files (CSS, JS)
• POST /transcribe - Transcription endpoint (proxy to Whisper API)
• POST /summarize - Summarization endpoint (proxy to LLM API)

File Structure

transcript-app/
├── Dockerfile         # Multi-stage Docker build with Red Hat UBI9
├── Makefile          # Build and run commands
├── README.md         # This file
├── prompt.md         # Application specification
├── server.go         # Go backend server
└── static/
    ├── index.html    # Web interface
    ├── style.css     # Styling
    └── app.js        # Frontend logic

Development

Local Development (without container)

If you want to run the Go server locally without containers:

# Install Go 1.23+
# Set environment variables
export AUDIO_INFERENCE_URL=http://localhost:8000
export AUDIO_MODEL_NAME=whisper-1  # Optional
export LLM_INFERENCE_URL=http://localhost:8001
export LLM_MODEL_NAME=gpt-3.5-turbo  # Optional

# Run the server
go run server.go

Testing with a Whisper API

You can use various Whisper API implementations:

1. faster-whisper: High-performance implementation
2. whisper.cpp: C++ implementation with HTTP server
3. OpenAI Whisper API: Official OpenAI API

Make sure your API is compatible with the OpenAI transcription endpoint format:

POST /v1/audio/transcriptions
Content-Type: multipart/form-data

file: <audio.wav>
model: whisper-1

Testing with an LLM API

You can use various LLM API implementations:

1. vLLM: High-performance LLM serving
2. Ollama: Local LLM with OpenAI-compatible API
3. LM Studio: Local LLM with API server
4. OpenAI API: Official OpenAI API

Make sure your API is compatible with the OpenAI chat completions endpoint format:

POST /v1/chat/completions
Content-Type: application/json

{
  "model": "gpt-3.5-turbo",
  "messages": [
    {"role": "system", "content": "You are a helpful assistant..."},
    {"role": "user", "content": "Summarize this text..."}
  ]
}

Technical Details

Backend (Go)

• Uses only Go standard library
• Handles multipart form uploads
• Proxies requests to Whisper API
• Serves static files efficiently

Frontend (JavaScript)

• Pure vanilla JavaScript (no dependencies)
• MediaRecorder API for audio capture
• WAV encoding for browser compatibility
• Modern async/await patterns

Container (Dockerfile)

• Multi-stage build with Red Hat UBI9 images
  • Build stage: registry.access.redhat.com/ubi9/go-toolset:1.23
  • Runtime stage: registry.access.redhat.com/ubi9/ubi-minimal:latest
• Handles UBI9 non-root user permissions correctly (UID 1001)
• Go 1.23 toolset for building
• Minimal runtime image for production
• No local Go development required

Browser Compatibility

• Chrome/Edge: ✅ Full support
• Firefox: ✅ Full support
• Safari: ✅ Full support (may need HTTPS for microphone)

Note: Microphone access requires HTTPS in production environments.

Troubleshooting

"AUDIO_INFERENCE_URL environment variable is required"

Make sure to set the required environment variables before running:

export AUDIO_INFERENCE_URL=http://your-whisper-server:8000
export LLM_INFERENCE_URL=http://your-llm-server:8001
make run

Microphone not working

• Check browser permissions for microphone access
• Ensure you're using HTTPS (required by most browsers)
• Try a different browser

Transcription fails

• Verify the Whisper API is running and accessible
• Check the API endpoint format is correct
• Review logs with make logs

Summarization fails

• Verify the LLM API is running and accessible
• Check that the LLM API endpoint supports the OpenAI chat completions format
• Ensure the LLM_URL is correctly set
• Review logs with make logs

File upload fails

• Ensure the file is in WAV format
• Check file size (limit: 500MB)
• Verify the API server is responding

License

MIT License - Feel free to use and modify as needed.

Contributing

This is a simple application built with standard libraries only. Feel free to fork and customize for your needs!