hive-voice-echo

A proof-of-concept voice loop: speak, get transcribed, get an answer, hear it back. Streamlit front end, Google Cloud Speech-to-Text and Text-to-Speech.

This was the first working end-to-end voice path in the hive — recorded audio in, synthesised audio out — built to find out where the loop actually breaks. It does break, and usefully: the failures were in encoding and in language switching, not in the models.

Running it

pip install streamlit google-cloud-texttospeech google-cloud-speech requests
export GOOGLE_APPLICATION_CREDENTIALS=/path/to/your-service-account.json
streamlit run app.py

The credentials path is read from the environment. There is no key in this repository and never was.

Files

• app.py — the Streamlit loop: record → transcribe → answer → speak
• synthesize.py — standalone text-to-speech helper
• enable_api.py — one-off API enablement

Licence

MIT.