Hacked together openai-whisper and deepl translation to stream and live transcribe-translate languages during meetings to any supported whisper/deepl languages.

The inspiration came from my struggle to understand some colleagues with a thick german/swiss german accent during meetings. This tool will hopefully help me follow along!

• Create Android APK package for smartphone
• Troubleshoot PyInstaller for MacOS dist (App is not performing as expected)
• Add App buttons for transcription source language and translation target language

• The App uses openai-whisper and deepl API to live transcribe and translate speech.
• Currently, the whisper model is hosted on a GPU server.
• Speech is recorded through the microphone and sent as WAV through WebSocket to /transcription API endpoint.
• Transcribed text is translated with deepl API to english.
• Translated text is streamed back to App.

• Clone the repository:

git clone [email protected]:K-Schubert/TransApp.git
cd TransApp

• Set env variables:

nano .env
DEEPL_API_KEY=<DEEPL_API_KEY>
TRANSCRIPTION_ENDPOINT=<TRANSCRIPTION_ENDPOINT_URL>

• Create venv:

# NOTE: Requires python < python3.13
python3.11 -m venv venv_transapp
source venv_transapp/bin/activate

• Install requirements:

pip install -r requirements.txt

• Run app and server with CLI:

# Run server
python3 whisper_server_stream.py

# Run app
python3 translate_app_stream.py

• OR build executable for MacOS:

# Build app
pyinstaller TransApp.spec
# Run app from dist folder