Ordering a burger with voice AI
After seeing and being impressed about how drive-thru personnel operate the conversation with the people in the cars, and at the same time prepare the food and coordinating with colleagues, I was curious on how to improve that process. So here s is AIBurger, a drive-thru experiment where you order by talking to an AI called Byte. You can ask what’s on the menu, interrupt, and change your mind. Your order builds on screen as you speak.
Here’s a short recording of me trying it. I order a meal, another burger, and fries, then ask to remove the pickles from one burger. After asking about dessert, I add a cookie and coffee and confirm the order.
Byte uses OpenAI’s realtime speech-to-speech model to listen and answer out loud. Every change to the order is a tool call checked against the actual menu. Adding a burger or removing its pickles needs to result in a valid change to the order.
Alongside the conversation, Mistral’s Voxtral transcribes what you’re saying. A small decision model, jev, uses those words to predict what you’re ordering in about 0.2 seconds, and the predictions appear as tentative items on screen until Byte confirms them. That gives the screen something interactive to show while the voice conversation catches up.
The same pieces could fit a kiosk, a counter, or a phone line. This is a small example of software that you can talk to while still having something concrete to look at.
You can try AIBurger with 10 free orders a day, or sign in with Opper to continue on your own account. There’s more detail on the project page. All three models run through Opper.