Whisper-GUI

A simple GUI made with gradio to use Whisper.

Requirements

• Anaconda or Miniconda installed and conda added to PATH.
• git installed and added to PATH. See instructions.
• ffmpeg installed and added to PATH. See instructions for Windows, Linux or macOS.

Optionally, to use GPU on Windows:

• CUDA version ≥12.0. Install from Nvidia's official site.

Set up

• In Windows, run the whisper-gui.bat file. In Linux / macOS run the whisper-gui.sh file. Follow the instructions and let the script install the necessary dependencies. After the process, it will run the GUI in a new browser tab.

Otherwise, manual steps are:

• Create a conda environment with Python 3.10
  conda create --name whisperx python=3.10
conda activate whisperx
• Install PyTorch 2.0
  For macOS:
  conda install pytorch::pytorch==2.0.0 torchaudio==2.0.0 -c pytorch
  For other OS, if you have GPU:
  conda install pytorch==2.0.0 torchaudio==2.0.0 pytorch-cuda=11.8 -c pytorch -c nvidia
  If not, for CPU:
  conda install pytorch==2.0.0 torchaudio==2.0.0 cpuonly -c pytorch
• Install whisperx and dependecies
  pip install git+https://github.com/m-bain/whisperx.git
  Original instructions in: https://github.com/m-bain/whisperX
• Install additional libraries
  pip install gradio
• Clone this repository
  git clone https://github.com/Pikurrot/whisper-gui

Run the GUI

To run the program every time, you can just run the same whisper-gui.bat or whisper-gui.sh (whatever your OS), which will also automatically check for updates of this repository.
Your transcriptions will be saved by default in the outputs folder of the repository.

Otherwise, to run manually:
conda activate whisperx
python main.py --autolaunch

Licensing

This project is primarily distributed under the terms of the MIT License. See the LICENSE file for details.

Third-Party Code
Portions of this project incorporate code from WhisperX, which is licensed under BSD-4-Clause license. This code is used in accordance with its license, and the full text of the license can be found within the relevant source files.