PlatePal: Food Scanner

Inspiration

Most food apps turn dinner into a math problem: search a database, weigh the chicken, log every calorie, then feel guilty about what's left. Most people give up within a week, and for some it hurts their relationship with food. We wanted something that feels more like a friendly nutritionist glancing at your plate and saying, "Nice portion. Maybe add some greens next time." That's how PlatePal started: one photo, clear guidance, no numbers, no judgment.

What it does

PlatePal turns a photo of your meal into simple, useful feedback:

📸 Snap your plate. No typing, searching or logging.
👕 See your portion size on a familiar XS–XL scale, based on your own daily needs.
⚖️ Check your plate's balance. It tells you whether the meal has protein, fiber and energy, and what it could use more of.
⚠️ Get gentle watch-outs when a plate is heavy on sugar or saturated fat.
🎯 Get advice that matches your goal, whether that's losing weight, maintaining, building muscle or just eating better.
🗓 Look back over time. A calendar of your plates shows your best meals and the habits worth noticing.

How we built it

PlatePal is a native iPhone app built with Swift and SwiftUI. The plate analysis uses Apple Intelligence's on-device Foundation Models, so on supported iPhones your photos are analyzed privately on the device. On other devices it falls back to OpenAI's vision models. Meals are saved to Apple Health (HealthKit), so your data stays yours and works with the rest of your health apps. iCloud keeps your meal photos in sync across devices.

Challenges we ran into

The hardest part was getting a camera to "see" a meal reliably. Estimating a portion from one photo is tricky, so we rewrote and retuned the AI many times. The second challenge was turning numbers into words: deciding when a plate counts as "medium" or "balanced" so it feels fair to everyone, whatever their goal. Supporting 18 languages, including the AI's own feedback, also meant making sure the app's friendly tone survives translation.

Accomplishments that we're proud of

A food app that gives genuinely helpful feedback without showing a single calorie count.
Private on-device AI analysis on supported iPhones.
Full integration with Apple Health.
Available in 18 languages from day one.

What we learned

The words you show people matter as much as the numbers behind them. "Hearty, could use some fiber" does more for a habit than "847 kcal, 6 g fiber." We also learned how capable on-device AI has become: private, fast analysis is now realistic on a phone.