/whispergo

Desktop ASR application powered by Whisper (and Wails)

Primary LanguageNSIS

WhisperGo

WhisperGo is an automatic speech recognition (ASR) application that transcribes audio files to text. It is powered by whisper.cpp, a C++ implementation of OpenAI's Whisper model. The application is built in Go and Svelte using Wails.

This project showcases the efficiency of the Whisper model, which is capable of transcribing audio files in real-time. The model is lightweight and can be run on a CPU, and larger versions of it can also be executed on a GPU.

Features

  • Efficient, real-time transcription of audio files.
  • Optional support for GPU acceleration.
  • Support for multiple Whisper models, varying in size and accuracy.
  • Clean and easy-to-use interface.

Demo

Prerequisites

  • Go 1.20+
  • Node.js 15+ and pnpm
  • Wails 2.8+ and its dependencies
  • GCC 12+, make and wget
  • (Optional) CUDA 12.0+ for GPU acceleration

Building

Clone the repository and navigate to the project directory:

git clone https://github.com/GarciaLnk/whispergo
cd whispergo
git submodule update --init --recursive

Build the project:

make

Run the application:

./build/bin/whispergo