CarbCam uses AI image recognition to identify the components of a meal, assign portion sizes and sum up the carbs based on USDA FoodData Central reference values. The estimate runs through three selectable AI engines (Gemini 2.5 Flash, Claude, OpenAI). Accuracy for normal plated meals is around ±15–25 %, slightly lower for mixed dishes or unknown brands. Use the AI calibration slider (50–100 %) to fine-tune the estimator to your own metabolism.

FPE stands for Fat-Protein Unit. One FPE equals 100 kcal from fat and protein combined — formula: FPE = (fat × 9 + protein × 4) ÷ 100. Meals high in fat and protein (e.g. pizza, burgers, rich sauces) often trigger a delayed blood-glucose rise over several hours that pure carbohydrate values miss. Pump users often use FPE values to configure an extended bolus (dual-/square-wave). CarbCam calculates FPE automatically and shows it as an amber chip alongside carbs (green) and kcal (blue).