Finished: AAC setup on a Pi 5 and XR glasses for my brother

Nov 19, 2026 Last reply: 13 hours ago 4 Replies
Jump to replies (4)

My brother controls everything with two switches mounted near his head, and up till now that meant a tablet on a mount. His screen app, Benny's Hub, is somthing I wrote for him myself, covering communication, games and streaming. The catch was that pressing a switch meant turning his head, so his eyes came off the tablet every time, which is rough given his vision. Plus fiddling with the mount position was a constant chore. When OpenAI picked us for thier Codex Physical Builds project I rebuilt things around it. Benny's Hub now runs on a Raspberry Pi 5 hooked up to VITURE XR glasses, input comes from two 3D printed switches, and a Bluetooth speaker carries his voice out loud. The glasses battery pack runs the Pi for over four hours, the lot packs into a small bag, and he gets a bit more privacy while using it too. That effort rippled back into the hub itself. There's now a lighter graphics mode so the 3D games run nicely on the Pi, a Chrome extension so he can watch Netflix, Disney+, Hulu and Plex then hop back to his own menu, and an option to send his voice out thru an external speaker. My other brother built him a custom voice maker as well, so he can have voices made just for him. Honest verdict: he still likes his tablet better for long sessions, glasses get tiring after hours. We still pull out the Pi setup periodically and it's been a great way to work out what suits him best. Everything is free and open source, build plans included, and it should work for anyone on one or two switches. Links in the comments.


The way you put it about the head turns pulling his eyes off the screen really stuck with me, because that tradeoff between switch position and line of sight is something I've wondered about watching my cousin use his single switch. Seeing it solved with a Pi and glasses rather than another mount tweak is genuinely inspiring. Your honesty about the glasses tiring after a long session makes the whole writeup more credible too, not less. I also want to say the nonprofit angle deserves its own applause. Using AI to help people like your brother, and making games for switch users in a community that gets almost nothing made for them, is the kind of thing coding should be for. For anyone hesitant reading this, the build plans and links being free and open source, and the whole thing working on one or two switches, means there's no barrier to trying it. Do you find the two switch layout changes which games work best, compared to what a single switch user would pick?

The part that hit home for me is the honest verdict at the end. My daughter is a single switch user and I spent a long time hunting for the one perfect setup, when really what we needed was the freedom to swap between two options depending on the day. Treating your Pi build and the tablet as tools to compare, rather than a replacement, seems like the healthier way round. The privacy angle resonates too. Anything that lets someone have a conversation without the whole room reading it over their shoulder feels like it matters more than we sometimes admit. One question from my side: with the Chrome extension letting him hop from streaming back to his own menu, is that transition instant or does he have to press through a couple of steps? For my daughter even one extra press can be the difference between using something and giving up on it. Also wanted to say the custom voice work is the detail I keep thinking about. A voice that actually sounds like him has to change how people respond to him, and that might be the biggest upgrade of the whole rebuild.

I'll push back on one bit of the 'no barrier to trying it' line though. Free plans and open source remove the cost, sure, but that's not the same as no barrier. You still need the hardware, the glasses, printed switches, and the patience to put it all together and tune it for whoever is using it. For a family already stretched thin, that's a real hurdle even with everything free. My own shelf is full of open source projects I never actually got round to building. Free gets you in the door, it doesn't walk you through it.

The custom voice maker your other brother built is the part that matters most to me here. Someone I support with her AAC gets frustrated that every voice option sounds like the same handful of presets, and she always perks up when something acutally fits her. Did you generate the voices from scratch or work from samples? The speaker output option makes sense to me too, a voice coming out of a device held right in front of you can feel a bit exposing.

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required