offline_sst

repo of files pertaining to realtime, offline translations using whisper realtime and argos translate. This repo is marked Creative Commons CC0. https://creativecommons.org/share-your-work/public-domain/cc0/

How it works: Whisper realtime produces text output which is piped into a watched file. A script watches this file and its triggering writes the output text into an executable command for Argos Translate. This executable is run and its output (translation) is echoed onto the terminal.

https://www.youtube.com/watch?v=gLVprfZREkI

Whisper real-time and Argos translate also work on Win and Mac but the tool 'inotifywait' is used for communication from one program to the other and is Linux specific. There are equivalent tools in those other operating systems but are not covered here. This is a work in progress.

Install dependencies

Anaconda3 portaudio19-dev ffmpeg setuptools-rust inotify-tools pulseaudio libasound-dev libportaudio2 libportaudiocpp0 alsa-utils libasound2-plugins

clone

github.com/autotunafish/offline_sst

conda env

conda create -n name-me python=3.9 conda activate name-me

build

cd offline_sst pip install -r requirements.txt

check run

python transcribe_demo.py --model tiny

• ctrl-C to stop

In a new terminal, Home, base env

pip install argostranslate argospm update argospm install translate

Add executable permissions to the scripts and receivingexec4.txt

Adjust the filepaths (Sorry!)

in Home, base env, run the script

./espres_o

in name-me, run

python transcribe_demo.py --model tiny 2>&1 | tee -a testing.txt

Check out

https://github.com/davabase/whisper_real_time https://github.com/argosopentech/argos-translate

Update: added line to print the translation with line breaks, much nicer.