A local web tool that analyzes food photos to estimate macros and calories using AI. No subscriptions, no manual logging, just data.

The Problem: The "Pay-to-Log" Fatigue

We’ve all been there: you want to track what you eat, you download a promising app, and before you can even search for "apple," you're hit with a "Start your 7-day free trial" popup.

Logging meals is already a chore; paying a monthly subscription just to access a database or use "AI features" felt wrong. I wanted a personal, straightforward tool that could tell me what’s on my plate without the friction of manual entries or paywalls. So, I built it.

The Solution: Nutrition Tracker

Nutrition Tracker is an experimental local application designed to bridge the gap between "taking a photo of my lunch" and "knowing my protein intake."

By leveraging Gemini Flash, the app performs a multimodal analysis of your food photos. It doesn't just guess; it identifies components and estimates portions to provide a breakdown of calories, proteins, carbohydrates, and fats in seconds.

How it works

1. Snap: Take a photo of your meal.
2. Analyze: The image is processed by the AI with a prompt optimized for nutritional density.
3. Results: You get an immediate estimate of your macros.

Tech Stack

• Gemini Flash: The brain of the operation. Chosen for its impressive speed and ability to "understand" visual volume better than previous models.
• React & Next.js: For a snappy, responsive interface.
• Tailwind CSS: Keeping the UI clean and minimalist.

A Note on Accuracy

Since this is a local, experimental project using AI for visual estimation, results should be taken with a grain of salt (pun intended). While Gemini is remarkably good at identifying ingredients, estimating exact weight from a 2D image is an art, not a perfect science. It’s meant to be a helpful guide, not a medical-grade tool.

I made this for myself, but I’m sharing it in case anyone else finds the "subscription-based-calorie-counting" world as exhausting as I do.