@whitequark✧✦Catherine✦✧

There’s an Android app that lets you run them stand alone as well.

It’s some really impressive work. The models for each language pair tend to be around 60 MB. They’re not good enough for high-stakes work (hire a translator!) but they’re great for a quick-and-mostly-okay translation. They’re trained from open corpora that are curated for building machine translation systems and they’re designed so that you can do the training on a single (powerful) machine. You can probably retrain one of them with a cloud VM for a few tens of dollars.

0

If you have a fediverse account, you can quote this note from your own instance. Search https://infosec.exchange/users/david_chisnall/statuses/116136334976564119 on your instance and quote it. (Note that quoting is not supported in Mastodon.)