Transcription Service App

Transcription Service App is a web app for universities to make simple transcriptions from video or audio files in multiple languages, currently tailored towards Open AI's Whisper models.

Some of its features are:

• Supports transcriptions with or without simultaneous translations to multiple languages.
• Simple interface.
• Configurable access to Open AI's Whisper models (tiny, base, small, medium, large-v1, large-v2, large-v3, large-v3-turbo).
• Supports upload from videos and audio files (up to 1gb).
• Users can edit and download transcription results in 4 different formats (txt, vtt, srt and json).
• Diarization support to detect multiple speakers (up to 20).
• Srt, vtt and json formats provide timestamp and speaker information (when available).
• Transcribed subtitles can be activated in uploaded videos.

Usage & Configuration

You first need to set up a whisperx API server to work with this app.

Some environment variables should be set. Here is an example of a .env file:

# PATH to the ffmpeg library in your system
FFMPEG_PATH=/usr/bin/ffmpeg
# Path where temporal files will be generated
TEMP_PATH=transcription-whisper-temp
# Uncomment this up if you're using an authentication process to allow users to log out
#LOGOUT_URL=/oauth2/sign_out
# Url and port to the API server
API_URL=http://111.111.111.11:11300
# Available Whisper models (comma-separated)
WHISPER_MODELS=tiny,base,small,medium,large-v1,large-v2,large-v3,large-v3-turbo
# Default model selection
DEFAULT_WHISPER_MODEL=base

# Prometheus metrics configuration
ENABLE_METRICS=true
METRICS_PORT=8000

Prometheus Metrics

The application includes comprehensive Prometheus metrics for monitoring usage and performance. When enabled, metrics are exposed on a separate HTTP endpoint for Prometheus to scrape.

Available Metrics

• Application Metrics:

  • transcription_page_views_total - Total page views by UI language
  • transcription_active_sessions - Number of active user sessions
  • transcription_app_info - Application version and metadata

• File Upload Metrics:

  • transcription_file_uploads_total - Total file uploads by type and status
  • transcription_file_upload_size_bytes - File upload size distribution

• Transcription Metrics:

  • transcription_jobs_total - Total transcription jobs by language, model, and status
  • transcription_duration_seconds - Transcription processing time distribution
  • transcription_active_jobs - Number of currently processing jobs
  • transcription_model_usage_total - Usage count per Whisper model
  • transcription_language_usage_total - Usage count per transcription language
  • transcription_speaker_detection_total - Speaker detection usage statistics

• User Interaction Metrics:

  • transcription_user_actions_total - User actions (save, edit) by format
  • transcription_downloads_total - File downloads by format

• API and Error Metrics:

  • transcription_api_request_duration_seconds - API request latency
  • transcription_errors_total - Error count by type and component

Configuration

Metrics are controlled by environment variables:

• ENABLE_METRICS=true - Enable/disable metrics collection (default: true)
• METRICS_PORT=8000 - Port for metrics HTTP server (default: 8000)

Prometheus Configuration

Add this job to your prometheus.yml:

scrape_configs:
  - job_name: 'transcription-app'
    static_configs:
      - targets: ['localhost:8000']
    scrape_interval: 15s
    metrics_path: /metrics

Grafana Dashboard

The metrics are designed to work well with Grafana. Key dashboard panels might include:

• Active sessions and transcription jobs over time
• Transcription success/failure rates
• Processing time percentiles by model
• File upload volume and sizes
• Language and model usage distribution
• Error rates and types

Testing Metrics

You can test the metrics functionality using the included test script:

python test_metrics.py

Development

The app is developed in the streamlit framework.

You can install the requirements needed to run and develop the app using pip install -r requirements.txt. Then simply run a development server like this:

streamlit run app.py

The metrics endpoint will be available at http://localhost:8000/metrics when the app is running.

Authors

virtUOS