Skip to content

This is an implementation of OpenAI's Whisper for the purpose of speech-to-text via default microphone, enabling direct output to clipboard.

License

Notifications You must be signed in to change notification settings

GucciFlipFlops1917/Whisper-Speech-To-Text

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

13 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Whisper-Speech-To-Text

This is an implementation of OpenAI's Whisper for the purpose of speech-to-text via your default microphone, enabling direct output to your clipboard and/or CLI. In addition, the script records and inferences with the press of a desired keystroke combination.

Installation

Installation is as easy as:

pip install -U openai-whisper

pip install -r requirements.txt

I also strongly encourage the installation of PyTorch with CUDA:

pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu117

If you choose to use PyTorch with your CPU instead, please run:

pip install torch

Usage

Most input arguments carry over from the base Whisper package.

Those of note are:


--model default is the 'small' model, some others are the 'tiny', 'base', 'medium', and 'large' models

--model_dir default is set as None, 'Users/[username]/.cache/whisper' is used as the default dir

--device default is 'cuda', set this to 'cpu' if you don't have cuda installed with torch

--task default is transcribe (duh), but translation is also possible

--language default is set as None (enabling language detection), but I recommend setting this to English/French/Spanish/etc. to cut down on inference times

--temperature default is 0, depolarizes output distribution allowing for more 'creativity'

--threads default is 0, this refers to CPU threads


Novel input arguments:


--amplification default is 1.0, amplifies the recording by a given floating point multiple

--print_text default is not set/False, prints text to your CLI/terminal

--no_clipboard default is not set/False, prevents output from being sent to your clipboard

--hotkey default is 'ctrl+shift+r', the combination that activates audio recording and inferencing

About

This is an implementation of OpenAI's Whisper for the purpose of speech-to-text via default microphone, enabling direct output to clipboard.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages