Need a bit more context. Which translation service and where on your phone are you auto translating?
Are you talking about Damus? Which version of Damus are you running? I have some changes queued up that will improve the general translation user experience and performance.
Seems like all the mentioned services (Microsoft, Google ML Kit, DeepL, LibreTranslate) support Japanese according to docs, or am I missing something? https://learn.microsoft.com/en-us/azure/cognitive-services/translator/language-support
LibreTranslate’s translations are generally poor in quality. I recommend DeepL if you want better quality translations and don’t mind KYC. Or if you do mind, wait a bit for me to implement more translation service options in Damus.
Thanks for letting me know! This feedback helps inform me about how to proceed with supporting more translation services.
Good feedback! I’ll pay more attention to the set of supported languages from each service.
Seems like Microsoft is online only. Have you tried DeepL? It’s quite good. I’m wondering how you compare the two.
Would you mind describing why you would use Microsoft’s over other translation services? I haven’t used it much so I would like to learn more. I value giving users choice, but also don’t want to add too much bloat to Damus code, which makes it harder to maintain.
Thanks for implementing Google ML Kit on Amethyst and paving the path there.
Heard about people’s struggles on Damus’ limited options on translation providers and translation quality. I will be adding Google’s ML Kit for offline translations (doesn’t seem to require sign up, so no KYC or capped quotas) and ChatGPT.
This is basically what you’re asking for. The tradeoff is that we add a dependency to Google, and requires storing translation models offline on disk and memory which can take a good amount of space. This is also what Amethyst does on Android. I’m happy to explore this translation approach as an option for Damus.
https://developers.google.com/ml-kit/language/translation/ios
I wish traveling by train was better in the US.
Not furious, just appreciative. Thanks!
It does not, unfortunately. I would be happy to implement it if it existed.
Sorry for eating into everyone’s DeepL translation quota on Damus. I have a pull request open. (1) Caches translations in memory (2) If auto-translate is off but a translation service is enabled in settings, it will still show the “Translate Note” button, even if the note isn’t selected. https://github.com/damus-io/damus/pull/843
Auto-translation caching incoming for next Damus build.
Oh, do you mean machine translations of people’s notes? Or the app strings? If the former, you need to go into the Damus in-app settings and enable a translation service like DeepL, sign up for a free API account, and paste the API key into Damus settings.
Thanks! I replied. No worries, it’s not the easiest concept and took me a bit to get a hang of it too.