Mastodawn

I wonder if anyone would like to make use of this new libretranslate instance I spun up with GPU acceleration. 60 requests per minute limit, unless you want an API key which will increase the limit. which I'm more than willing to give out for free for any other instance who might want a translation service.

You can find it at

https://translate.bark.lgbt/

to test it out for yourself

#fediadmin #fedi #translate

LibreTranslate - Free and Open Source Machine Translation API

Free and Open Source Machine Translation API. Free to download, offline capable and easy to setup. Run your own API server in just a few minutes.

Show thread

Stekke Mar 25

@dalite I tried libretranslate a few months ago, but sadly found the quality of the translation to be quite poor. If you want high quality local translations and can wait a few seconds for it then translategemma is excellent.

Show thread

Dalite Fur 🦊Mar 25

@bat That's actually quite interesting! I've ran gemma via ollama and openwebui but I didn't know they made a model specific for translating. I don't think it integrates with mastodon though as that one only lists support for DeepL and libretranslate api's

Show thread

Stekke

@dalite yeah. I had quite a lot to translate, and libretranslate was my first attempt. But it just messed up a lot of translatios, I couldn't trust it. Large llm's are too slow for such a task. But gemma3 is still reasonable fast for non-realtime translations. For the translategemma they give you a very specific short prompt you have to follow.

Idk about any integrations for it though, like most of the AI stuff it seems like you kinda need to do your own plumbing.