Whisper Live (Offline Speech Recognition)

This project uses speech_recognition and faster_whisper to record live speech and transcribe it into text.

How to Install

Exe	Description	Releases
	Download the latest version	Download

Build from Source

git clone https://github.com/NxRoot/whisper-live.git

Create a Virtual Environment

Win32

python -m venv venv

MacOS

python3 -m venv venv

Activate the Virtual Environment

Win32

.\venv\Scripts\activate

MacOS

source venv/lib/activate

Install Required Modules

Win32

pip install -r requirements.txt

MacOS

pip3 install -r requirements.txt

Test the Application

Win32

python main.py

MacOS

python3 main.py

Build into Executable

Make sure you have PyInstaller in your virtual environment.

Win32

pyinstaller --name wl --onedir main.py --icon icon.ico --add-data "model;model"

You can use the --onefile argument instead of --onedir to build a portable version.

MacOS

pyinstaller --name wl --onedir main.py --icon icon.ico --add-data "model:model" --paths=./venv/lib/python3.10/site-packages

Make sure the --paths argument is aligned with your python version.

 

Custom Execution

This can be executed from a terminal or called from another app.

Default

wl.exe

Show Info

wl.exe --verbose

Debug Logs

wl.exe --verbose > logs.txt

Call Examples

Make sure you place wl.exe in the same folder as index.js

NodeJS

const { join } = require("path");
const { spawn } = require("child_process");

const ps = spawn(join(__dirname, "wl.exe"), ["--verbose"]);

ps.stdout.on("data", (data) => {
  console.log(`${data}`.trim());
});

ps.stderr.on("data", (data) => {
  console.error(`${data}`.trim());
});

Important Notes

• Always clone the repo if you wish to build from source.
• The model can take some time to load using portable versions.
• On MacOS the SpeechRecognition module requires PortAudio installed.

 

⭐ If you find this useful!