Skip to content

Almost online speech translation on Apple Silicon laptops with CoreML enabled. Doesn't need any APIs, all work is done locally using OpenAI's excellent Whisper model. Also https://github.com/ggerganov/whisper.cpp repo is used to build Whisper with CoreML support, enhancing speed significantly

Notifications You must be signed in to change notification settings

AndreyKolomiets/local_speech_translator

Repository files navigation

usage: ./run_all.sh <output folder> <initial prompt>

need to fill .env file first with the following variables:

DEFAULT_INPUT_DEVICE_NAME = <input device name to use as default>
EXECUTABLE_PATH = <path to pre-built whisper.cpp executable>
MODEL_PATH = <path to whisper weights>

jupyter notebook will list all the devices available

ui will start at http://127.0.0.1:8050/

Tested on python 3.10

About

Almost online speech translation on Apple Silicon laptops with CoreML enabled. Doesn't need any APIs, all work is done locally using OpenAI's excellent Whisper model. Also https://github.com/ggerganov/whisper.cpp repo is used to build Whisper with CoreML support, enhancing speed significantly

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published