This Python script allows users to record audio using their microphone and transcribe the recorded audio using the Whisper transcription model. It provides a simple hotkey to start and stop recording using a customizable hotkey combination.
- Start and stop recording using a hotkey combination
- Transcribe recorded audio using the
Whispermodel from OpenAI (free and offline) - Copy the transcription to the clipboard automatically
- Python 3.7, 3.8, 3.9
- CUDA is highly recommended for performance
- Install
PyTorchfor python. Refer to the PyTorch's website for details - Clone this repository and install the requirements
git clone https://github.com/Crossoufire/whisper2clip.git cd whisper2clip pip install -r requirements.txt - Install
Ffmpegand add it to the PATH (necessary for the Whisper model)
- When executing this script for the first time will download the chosen Whisper model (
default: medium.enabout1.4Gb) - Based on your GPU 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, the
.enmodels (e.g.,tiny.en,base.en) tend to perform better - To change the model, modify these lines in the
main()function of themain.pyscript
hotkey = "ctrl+alt+space"
model_name = WhisperModels.MEDIUM_EN
- The available
WhisperModelsare the same as presented in the table above as Python Enum.
- Run the application
python main.py
- Enter the hotkey to start recording (default:
ctrl+alt+space) and again to stop recording - Transcription will be automatically copied to your clipboard
- This project uses the OpenAI's Whisper model for the audio transcription
- This project is forked and modified from the original creator gustavostz