Whisper Web UI

A browser interface based on the Gradio library for OpenAI's Whisper model.

Features

• Transcribes videos from YouTube, a video or audio file or a recording from your microphone.
• Output
  • Text
  • Markdown with Timestamps
  • JSON with the text, start and end time
  • JSON with all parameters
  • Downloadable as file
• The system default dark theme is currently not supported in all browsers.
  • To switch to dark mode manually add /?__theme=dark to the end of the URL.

Hardware Requirements

You can run this on your CPU but it is a lot slower than running it on an GPU.
See this for the GPU hardware requirements.
The models tiny, base and small should run on any hardware.

Installation and Running

Installation on Windows

1. Install Python 3, checking "Add Python to PATH"
2. Install git
3. Install ffmpeg
4. Reboot
5. Download this repository, for example by running git clone ...........
6. (optional) To use your GPU install PyTorch manually (see Use a GPU)
7. Install the required Python dependencies with pip3 install -r requirements.txt or pip install -r requirements.txt
8. Copy the .env.example file and rename the copied file to .env
9. (optional) Change the parameters in the .env file
10. Run py app.py

Installation on Linux

1. Install dependencies:

   # Debian-based:
   sudo apt update && sudo apt install git python3 python3-venv python3-pip ffmpeg
   # Red Hat-based:
   sudo dnf install git python3 python3-pip ffmpeg
   # Arch-based:
   sudo pacman -S git python3 python-pip ffmpeg

2. Download this repository, for example by running git clone ...........
3. (optional) To use your GPU install PyTorch manually (see Use a GPU)
4. Install the required Python dependencies with pip3 install -r requirements.txt or pip install -r requirements.txt
5. Copy the .env.example file and rename the copied file to .env
6. (optional) Change the parameters in the .env file
7. Run python3 app.py or python app.py

Use a GPU

To use it with a GPU, install Nvidia's CUDA or AMD's ROCm drivers.
To install PyTorch with the right driver version see here: https://pytorch.org/get-started/locally/
Example: To install PyTorch for CUDA version 11.6: pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu116

Credits

• OpenAI Whisper https://openai.com/blog/whisper/, https://github.com/openai/whisper