whisper-write

TL;DR - Simulate keyboard typing with voice commands on your computer. Blazingly fast transcription is now a reality!⚡️ Whisper words in your computer.

Video demo:

Simulate keyboard typing with voice commands on your computer with OpenAI's Whisper Large v3.

Run Locally (whisper.cpp)
or on the cloud (OpenAI API)

Start the wkey listener. Keep a button pressed (by default: right ctrl_l) and speak. Your voice will be recoded locally. When the button is released, your command will be transcribed via Whisper and the text will be streamed to your keyboard.

You can use your voice to write anywhere. Now with support for whisper.cpp to locally transcribe your voice.

If you use Whisper API, you will incur costs. Currently, it costs $0.36 for 1 hour of transcription.

Setup

Install the package.

pip install -e .

You will need to set a few environment variables:

WKEY: the keyboard key you want to use to start recording. By default, it is set to right ctrl. You can use any key. Note that Mac and Windows might have different key codes. You can run fkey to find the code of the key you want to use.

If you plan on using OPENAI's Whisper API instead of local model, you will also need to set:

OPENAI_API_KEY: your personal OpenAI API key. You can get it by signing up here: https://platform.openai.com/

You can set the environment variables in your shell:

export OPENAI_API_KEY=<your key>
export WKEY=ctrl_l

Usage

Run wkey in a terminal window to start listening. If you prefer to use the Whisper API instead of the local model, run wkey --api instead.

Run fkey to find the code of the key you want to use.

If there are issues, check the additional requirements.

Planned features

CoreML build version of whisperCPP, ref

Build model guide ref
WhisperCPP Direct API Binding to use such model ref
- Since WhisperCPP doesn't yet support directly loading CoreML build models.

Additional requirements

Requirements differ depending on your OS.

Ubuntu

You will need to install the portaudio library.

sudo apt-get install portaudio19-dev

Mac

You will need to authorize your terminal to use the microphone and keyboard. Go to System Settings > Privacy and Security. Then:

Select Microphone and authorize your terminal.
Select Accessibility and authorize your terminal.

Restart the terminal for the changes to take effect.

Note that this might entail security risks.

Windows

Haven't tested it on Windows yet. If you do, please let me know how it goes.

Distribution as an executable

You can distribute the script as an executable. This will allow you to run it without having to install the package. To do so, run:

pip install pyinstaller
pyinstaller --onefile scripts/wkey

and the open the executable in the dist folder.

Security risks

This script creates a recording with your microphone and sends the audio to the Whisper API. The Whisper API response will be automatically streamed to your keyboard and executed there. This might entail security risks. Use at your own risk.

tshrjn / whisper-writer