Meet Gemma Translator: a completely offline, DIY translation device powered by Gemma 4 and built with Antigravity. 🧵

🛠️Hardware All the parts are off the shelf and easy to find: a Raspberry Pi 5, portable battery pack, tiny touchscreen display, mini speaker, and micro keyboard with a knob to select languages. All held together by custom 3D printed housing.

🧠Brain We use LiteRT-LM to run Gemma (gemma4-e2b) on-device with a simple Python server to handle requests. All the translation knowledge is built right into the model: no APIs or wifi required.

🗣️Hearing and speaking Both speech recognition and text-to-speech run on-device too. We used Moonshine, an open source model which handles ASR for lots of languages. Moonshine comes with Kokoro, a small model which we use to turn Gemma’s output text into speech.

👉UI We built a simple web frontend in React with a monochrome theme that shows info like language selection and mic feedback. It handles the hardware controls like knob turns and push-to-talk as keyboard events.

🔧 DIY You can build it all yourself by cloning our repo: You could run this on a Raspberry Pi, or even your laptop or Android device. That’s the beauty of small, on-device models like Gemma. If you want to add more features, try loading the project in Antigravity and telling it what to do.

Who even wants a local translator on a raspberry pi with a screen if you can just do it on your phone?

Wanted to build exactly same device , good to see someone do it

3.5 Pro wen

oh that's cool, I want to build cool things like this oh wait I can I just need robotic parts budgets

Antigravity please make a version of Google eloquent to android

??????????? when 3.5 pro

A fully offline translator is such a clever DIY build.

@grok What's the budget for setting all this up?

Finally, a translator that can survive airplane mode, bad reception, and the eternal search for a bathroom.

🔥🔥

@grok how are the limits in Antigravity for the $20 plan and how does it compare to competition (Codex, Claude, Grok and Cursor) Which one do you recommend?

The fact that people can build their own translator at home is wild.

Could yall improve antigravity desktop so its like Claude desktop? That would be insane. Would love to get all the quality and features claude desktopp has

Very cool! Offline AI translation is a huge step toward making language tools more private and accessible. 🚀

Based on my experience building similar devices, it needs at least Whisper V3 Turbo to handle speech recognition well in auto mode and other complicated things. Sorry, but this device looks like cute toy for me))

please antigravity, can you post a video of the production of this in a blog format so i can do similar things? especially the 3d modeling part

offline translation with no cloud dependency is the real win here, that speaker sketch on the notepad looks like the actual prototype stage.

Pushing the button to speak is not very ux friendly

I have no idea how people are building anything with gemini 3.6 flash when it struggles with the most basic questions. 2+2=? Gemini be like: I'm sorry, I can't generate this image, can I help you with anything else?

The physical language knob is the best part: it still works with gloves, no signal, and no app menus.

antigravaty just delete my windows😂😂😂

Did antigravity also do the 3d model / stl file?

cool!

waiting for gemini 3.6 pro ..

When will you release 3.5 😭😭

😀

woah google s best new ai model is so cool

💯💯

Awesome stuff It's so amazing to see what people are doing. So wild to live now and see all those fantastic things happening

I wake up, open Antigravity, the chat from yesterday night. Gone. Again. It has happened a few times. It's a text interface... Is the stock going down? Good.

Heres a more detailed article on how to configure antigravity cli subagents, and orchestration: