translatelocally features:
github.com
It works with your gpu and translates as you type
but it only has a limited number of languages
you can add extra repositories to provide more languages but i have only found 3
My freebsd-nixpkgs project allows you to install application
with the nixpkgs package manager in a Podman container with cuda support
freebsd-nixpkgs project
github.com
translatelocally install
github.com
I have only managed to find these language model links
Bergamot: https://translatelocally.com/models.json
OpusMT: https://object.pouta.csc.fi/OPUS-MT-models/app/models.json
HPLT: https://raw.githubusercontent.com/hplt-project/bitextor-mt-models/refs/heads/main/models.json
Bergamot is the default and you have to add the other two
translateLocally use the same engine as Firefox uses for local translation
So in theory you should be able to get the links to the models Firefox use's for translation
and then use them with translateLocally
There is a Firefox models.json file but it not compatible with translateLocally
because it not in the same format as the above links
I wasted a couple of hours trying to get Gemini to create a python script to convert the firefox models.json file
into the correct format for translateLocally but it totally failed
I had to repeat the task every 2 minutes over 20 times before i gave up
Converting the Firefox models.json file into the correct format for translateLocally
would enable you to have a locally hosted translation powered by the gpu
Using either the gui or on the command line
Instead of using google translate on the web
or the translate-shell command line tool which use google and bing in the backend
Trying to convert the firefox models.json file which i think has 143 languages by hand would be a pain
I would have to try one language and then see if the format is correct and actually works
Before working my way through the file
But it would be much easier if it was possible to use python or something to convert file
Im British and we have mastered an advanced linguistic technique
that allows us to communicate with anyone in the world without knowing their language
Its called point and shout
But people in other countries havent mastered the technique yet
so having local translation could be useful for them
GitHub - XapaJIaMnu/translateLocally: Fast and secure translation on your local machine, powered by marian and Bergamot.
Fast and secure translation on your local machine, powered by marian and Bergamot. - XapaJIaMnu/translateLocally
github.com
It works with your gpu and translates as you type
but it only has a limited number of languages
you can add extra repositories to provide more languages but i have only found 3
My freebsd-nixpkgs project allows you to install application
with the nixpkgs package manager in a Podman container with cuda support
freebsd-nixpkgs project
GitHub - NapoleonWils0n/freebsd-nixpkgs: Nixpkgs Podman containers on Freebsd with home-manager and flakes
Nixpkgs Podman containers on Freebsd with home-manager and flakes - NapoleonWils0n/freebsd-nixpkgs
github.com
translatelocally install
GitHub - NapoleonWils0n/freebsd-nixpkgs: Nixpkgs Podman containers on Freebsd with home-manager and flakes
Nixpkgs Podman containers on Freebsd with home-manager and flakes - NapoleonWils0n/freebsd-nixpkgs
github.com
I have only managed to find these language model links
Bergamot: https://translatelocally.com/models.json
OpusMT: https://object.pouta.csc.fi/OPUS-MT-models/app/models.json
HPLT: https://raw.githubusercontent.com/hplt-project/bitextor-mt-models/refs/heads/main/models.json
Bergamot is the default and you have to add the other two
translateLocally use the same engine as Firefox uses for local translation
So in theory you should be able to get the links to the models Firefox use's for translation
and then use them with translateLocally
There is a Firefox models.json file but it not compatible with translateLocally
because it not in the same format as the above links
I wasted a couple of hours trying to get Gemini to create a python script to convert the firefox models.json file
into the correct format for translateLocally but it totally failed
I had to repeat the task every 2 minutes over 20 times before i gave up
Converting the Firefox models.json file into the correct format for translateLocally
would enable you to have a locally hosted translation powered by the gpu
Using either the gui or on the command line
Instead of using google translate on the web
or the translate-shell command line tool which use google and bing in the backend
Trying to convert the firefox models.json file which i think has 143 languages by hand would be a pain
I would have to try one language and then see if the format is correct and actually works
Before working my way through the file
But it would be much easier if it was possible to use python or something to convert file
Im British and we have mastered an advanced linguistic technique
that allows us to communicate with anyone in the world without knowing their language
Its called point and shout
But people in other countries havent mastered the technique yet
so having local translation could be useful for them