🎙️ Virtual Voice Assistant

A simple voice assistant powered by ElevenLabs Conversational AI. It listens via your microphone, transcribes speech, and responds with synthesized voice.

🧩 How it works (visual)

flowchart LR
   U[You 🎤] -->|speak| MIC[Microphone]
   MIC --> E[ElevenLabs Conversational AI]
   E -->|STT| T[Transcript 📝]
   E -->|LLM| R[Response 💬]
   E -->|TTS| S[Audio 🔊]
   S --> SPK[Speakers]

Loading

✅ Setup

1. Create an environment file locally (do not commit secrets):

   • Copy .env.example to test.env
   • Fill in values for AGENT_ID and API_KEY

2. Install dependencies:

pip install -r requirements.txt

Alternatively (quick install, not pinned):

pip install "elevenlabs[pyaudio]" python-dotenv

Note: Voice input/output relies on PortAudio.

• Linux:

sudo apt install portaudio19-dev

• macOS:

brew install portaudio

3. (Windows) If PyAudio install fails, try:

pip install pipwin
pipwin install pyaudio

▶️ Run

python -u voice_assistant.py

Say “stop”, “exit”, “quit”, “bye”, “goodbye”, or “end session” to end the conversation.

⚙️ Configuration

Non-secret text is hardcoded in voice_assistant.py:

• user_name: default is "Robil"
• prompt: default is "You are a helpful assistant."
• first_message: default greeting using user_name

You can change these lines directly if you want different behavior. Secrets (AGENT_ID, API_KEY) stay in test.env.

🔐 Safety & Git Hygiene (if you commit)

• Do not commit test.env or any .env files. They are excluded by .gitignore.
• Logs like conversation_log.txt are ignored.
• If a key is accidentally committed, rotate it in the ElevenLabs dashboard immediately.

🤖 ElevenLabs Agent Setup

• Create an agent in the ElevenLabs dashboard (Agents → New agent).
• In the agent editor:
  • Set your System prompt and First message as desired.
  • Open Security and enable Overrides for at least: First message and System prompt (so runtime overrides from the app are allowed).
  • Click Publish to apply changes.
• Create an API key (Settings → API Keys) with permissions including ElevenLabs Agents Write (required for Conversational AI). Grant Text to Speech / Speech to Text access if you use audio.
• Copy the Agent ID into AGENT_ID and the API key into API_KEY in your local test.env.
• Ensure your account plan includes access to Conversational AI features.

🙏 Credits

This project was inspired by the Codédex tutorial “Create a Voice Virtual Assistant with ElevenLabs” by Alexandre Sajus:

• https://www.codedex.io/projects/create-a-voice-virtual-assistant-with-elevenlabs

I implemented my own version and added small improvements (logging, safer repo setup, and voice-session exit keywords).

📄 License

This project is licensed under the MIT License. See LICENSE for details.