Whisper Web

Whisper-web is a webapplication that allows you to transcribe sound files to text completely locally in your web browser.

This repository is a fork of Xenova/whisper-web.

Here are the main differences:

• Actively maintained
• Up-to-date dependencies, including transformers.js
• Ability to use WebGPU or CPU
• More user-friendly interface
• User interface in several languages
• Available as a progressive web app (so usable offline if added to your homescreen)
• Transcription is rendered continuously and not at the end
• Export to SRT
• Choose between a larger range of models (for example Swedish and Norwegian finetunes from the countries' national libraries)
• Choose your own quantization level for the model
• Clear cache with a button

The main application is available at whisper-web.mesu.re. It is hosted on Github Pages.

KB-Whisper

Initially, this project aimed at making the Swedish KB-Whisper models fine-tuned by the Swedish National library ♥️ more available for easy transcription of Swedish audio.

A version of the website with Swedish as default language is still available at kb-whisper.mesu.re (hosted in the EU by statichost.eu) and the source code is on the swedish branch but it is identical to the other version at whisper-web.mesu.re.

Running locally

1. Clone the repo and install dependencies:

   git clone https://github.com/PierreMesure/whisper-web.git
   cd whisper-web
   npm install

2. Run the development server:

   npm run dev

3. Open the link (e.g., http://localhost:5173/) in your browser.