My partner and I speak different languages, and there were no good options available for the macOS that did what I wanted.
I wanted something simple, always available, fast, offline and most importantly designed to show another person the translation from across the room.
This is what I came up with, should work with most Macs. You can use the in-built translation by Apple, use Apple Intelligence if you have it available, or you can use the local in-built LLM for translation (Google TranslateGemma powered llama.cpp)
Report
@butttons Congratulations on the launch! As somebody who is also in a multi-lingual partnership, I can definitely see the benefit of this. It's great that you took your life experience and translated it (no pun intended) into something actionable!
Has the product had a noticeable effect on the communication between you two?
@michael_nash2 Oh 100%. It's too early to say, but I'm constantly using it and it is noticeably faster than doing the usual ritual of normal browser translations.
The different models also give different kinds of results. Some give more professional responses, some give more casual ones and it's great to ask my partner the differences.
Report
On-device translation will hit scale pain on model footprint and latency across older Intel Macs, plus handling long text with consistent terminology.
Best practice is a tiered pipeline: use Apple Translation/Apple Intelligence when available, otherwise run a quantized local model with streaming chunks and a small glossary cache for repeated phrases.
How are you choosing between Apple frameworks vs llama.cpp at runtime, and do you plan per-language downloadable packs with offline quality benchmarks?
@ryan_thill users can chose between Apple intelligence and llama.cpp from the settings. Same for disabling AI altogether and using the language packs. There's also a switch for quickly toggling AI on/off in the main translation view.
Honestly, this is primarily intended for smaller sentences, so I haven't considered scaling issues for long texts. I did hit issues with longer texts with apple intelligence, that's why I added the llama.cpp models. Those seem to work fine.
Report
I really like the idea of being able to display translations across the room. That adds a unique, shared experience to it. My first thought was how this differs from Mac’s built-in translation tool, but that feels more geared toward personal use. This seems to introduce a more in-person, collaborative dynamic that sounds genuinely convenient.
@lunarturtle pretty much, Kyle.
The in-built tool is great for quick use 100%. But it's kind of awkward with the whole spotlite type inline translation.
Report
I like this app. Is it possible to set shortcuts without having to copy and paste content into Plea? I tried to contact via email, but the email bounced back. Please check if your email is working.
Report
On-device translation is the way to go. I run a multilingual app (8 languages) and privacy + speed matter so much. Love that it's a one-time purchase too.
FlareBar
@butttons Congratulations on the launch! As somebody who is also in a multi-lingual partnership, I can definitely see the benefit of this. It's great that you took your life experience and translated it (no pun intended) into something actionable!
Has the product had a noticeable effect on the communication between you two?
FlareBar
@michael_nash2 Oh 100%. It's too early to say, but I'm constantly using it and it is noticeably faster than doing the usual ritual of normal browser translations.
The different models also give different kinds of results. Some give more professional responses, some give more casual ones and it's great to ask my partner the differences.
On-device translation will hit scale pain on model footprint and latency across older Intel Macs, plus handling long text with consistent terminology.
Best practice is a tiered pipeline: use Apple Translation/Apple Intelligence when available, otherwise run a quantized local model with streaming chunks and a small glossary cache for repeated phrases.
How are you choosing between Apple frameworks vs llama.cpp at runtime, and do you plan per-language downloadable packs with offline quality benchmarks?
FlareBar
I really like the idea of being able to display translations across the room. That adds a unique, shared experience to it. My first thought was how this differs from Mac’s built-in translation tool, but that feels more geared toward personal use. This seems to introduce a more in-person, collaborative dynamic that sounds genuinely convenient.
FlareBar
I like this app. Is it possible to set shortcuts without having to copy and paste content into Plea?
I tried to contact via email, but the email bounced back. Please check if your email is working.
On-device translation is the way to go. I run a multilingual app (8 languages) and privacy + speed matter so much. Love that it's a one-time purchase too.
FlareBar