aspctu/substrate

on-device, real-time voice transcription

substrate

installation

First, clone this repo and install whisper.cpp in this directory

git clone https://github.com/aspctu/substrate.git
cd substrate

git clone https://github.com/ggerganov/whisper.cpp
cd whisper.cpp
make

Second, install other reqs

brew install portaudio ffmpeg
poetry install

usage

poetry shell
python3 main.py

Transcriptions will land in output.txt in real-time unless invoked with output flag

python3 main.py --output="path/to/my/file"

using other whisper models

By default, this uses the base.en model. You can use smaller / larger models as you wish.

First, download the model you want

cd whisper.cpp
make [size you want, e.g large]

Second, update the variable WHISPER_MODEL in constants.py

- WHISPER_MODEL = "base.en"
+ WHISPER_MODEL = "large"

credits

❤️ to whisper.cpp and speech_recognition