Post

Axel Le Pennec

@alpennec.com

Is there a local LLM I can run on my Mac to translate my app’s string catalogs? #xcode #iosdev

6/5/26, 6:34 pm1 repost2 likes

Dirk

@dirkolbrich.de

6 May

If you can split it up into small chunks, maybe apfel.franzai.com can be useful. Otherwise for larger local models I use pi.dev as an agent and oMLX.ai for the model server.

apfel.franzai.comapfel - Your Mac Already Has AIYour Mac already has AI. Apple ships a language model with macOS -apfel unlocks it with one brew install. No downloads, no API keys, no config. The fastest path to local AI.

Dirk

@dirkolbrich.de

6 May

Oh yeah, just saw translate.franzai.com which also uses the build in macOS model.

translate.franzai.comtranslate - Apple Translation from your terminalUNIX filter + drop-in HTTP server for DeepL, LibreTranslate, and Google v2. 100% on-device, no cloud, no API keys.

Axel Le Pennec

@alpennec.com

7 May

Thank you. I played with an app that uses the Apple Translation API but the results were not as good as I hoped notably because you can’t (if I’m not wrong) provide any context about the translation (like I could provide EN + FR + a comment). So I think an AI LLM would work better for this task.

Axel Le Pennec

@alpennec.com

7 May

I’ll check but I think I underestimated the power required to run local models or overestimated the performances of my M2 Pro. Or I haven’t found the good model for this translation task because I played with Osaurus and it was very very slow and unusable unfortunately.

Dirk

@dirkolbrich.de

11 May

Someone just made a CLI using the underlying Foundations framework for translating complete string catalogs. Maybe try this. github.com/JoshuaSulliv...

github.comGitHub - JoshuaSullivan/local-localizer: A command line app that uses Apple Intelligence to translate strings files.A command line app that uses Apple Intelligence to translate strings files. - JoshuaSullivan/local-localizer

Axel Le Pennec

@alpennec.com

11 May

Thanks for sharing! I’ll have a look but I’ll first have to update to Tahoe to try it 😆 In the meantime, I build a small app on my side that uses Gemma 4 + LM Studio. Not perfect yet but it gives promising results.

Matt Waller

@wattmaller.bsky.social

8 May

Might be able to find a good balance after getting Ollama up and running and trying a bunch of different models? You'll definitely want to break up your strings into individual objects as well, so that it's not tackling them all at once. Maybe Gemma 4?

Axel Le Pennec

@alpennec.com

9 May

I tried Ollama with Gemma 4 but I switched to LM Studio with Gemma 4. I enjoy the UI most. So far I tried to translate the whole JSON into a given language but it’s not providing good results notably because the format is not totally respected. I’ll split and try an API approach w/ individual calls.