ChatGPT Voice Assistant

Talk to an OpenAI model using your microphone. Speech is transcribed by OpenAI and the answer is spoken using OpenAI, Apple, or Google text-to-speech. API usage incurs charges.

Install

Install uv and Git. Python 3.12 is recommended. On macOS, install PortAudio with brew install portaudio. On Debian/Ubuntu, install it with sudo apt-get install portaudio19-dev python3-dev. A microphone and audio output are required; grant your terminal microphone permission on macOS.

Then install the CLI from this repository:

curl -fsSL https://raw.githubusercontent.com/jakecyr/chatgpt-voice-assistant/master/install.sh | bash

The script installs the current repository version with uv tool install. Run uv tool update-shell and restart your shell if gptassist is not on your PATH. You can inspect install.sh before running it. The URL uses the published master branch, so changes in a local checkout must be published before this command includes them.

Run

export OPENAI_API_KEY="your-api-key"
gptassist

Choose a microphone if prompted, then speak. Press Ctrl+C to stop. For options run gptassist --help. For example:

gptassist --open-ai-model gpt-6.1-terra --reasoning-effort low --max-tokens 1000 --tts openai

The default chat model is gpt-6.1-terra with low reasoning effort. The completion token limit includes reasoning tokens; raise it if replies are cut short. Model availability and supported reasoning levels depend on your OpenAI account and model. Transcription uses whisper-1; OpenAI speech uses tts-1. Google speech output uses network access; Apple speech requires macOS. --lang and --tld configure Google speech output. Never share your API key.

From source

git clone https://github.com/jakecyr/chatgpt-voice-assistant.git
cd chatgpt-voice-assistant
uv sync --locked
uv run gptassist

See DEVELOPMENT.md for tests and CONTRIBUTING.md for pull requests and releases.