airhors/whisper-clip

用語音找圖

WhisperClip: One-Click Audio Transcription

WhisperClip simplifies your life by automatically transcribing audio recordings and saving the text directly to your clipboard. With just a click of a button, you can effortlessly convert spoken words into written text, ready to be pasted wherever you need it. This application harnesses the power of OpenAI's Whisper for free, making transcription more accessible and convenient.

Table of Contents

• Features
• Installation
  • Prerequisites
  • Setting Up the Environment
  • Choosing the Right Model
• Usage
• Configuration
• Feedback
• Acknowledgments

Features

• Record audio with a simple click.
• Automatically transcribe audio using Whisper (free).
• Option to save transcriptions directly to the clipboard.

Installation

Prerequisites

• Python 3.8 or higher
• CUDA is highly recommended for better performance but not necessary. WhisperClip can also run on a CPU.

Setting Up the Environment

1. Clone the repository:

   git clone https://github.com/gustavostz/whisper-clip.git
   cd whisper-clip
   

2. Install PyTorch if you don't have it already. Refer to PyTorch's website for installation instructions.

3. Install the required dependencies:

   pip install -r requirements.txt
   

Choosing the Right Model

Based on your GPU's VRAM, choose the appropriate Whisper model for optimal performance. Below is a table of available models with their required VRAM and relative speed:

Size	Required VRAM	Relative speed
tiny	~1 GB	~32x
base	~1 GB	~16x
small	~2 GB	~6x
medium	~5 GB	~2x
large	~10 GB	1x

For English-only applications, .en models (e.g., tiny.en, base.en) tend to perform better.

To change the model, modify the model_name variable in config.json to the desired model name.

Usage

Run the application:

python main.py

• Click the microphone button to start and stop recording.
• If "Save to Clipboard" is checked, the transcription will be copied to your clipboard automatically.

Configuration

• The default shortcut for toggling recording is Alt+Shift+R. You can modify this in the config.json file.
• You can also change the Whisper model used for transcription in the config.json file.

Feedback

If there's interest in a more user-friendly, executable version of WhisperClip, I'd be happy to consider creating one. Your feedback and suggestions are welcome! Just let me know through the GitHub issues.

Acknowledgments

This project uses OpenAI's Whisper for audio transcription.