Point, type, or scan a barcode.
Three ways into the same log: a photo of the plate, a typed description, or a packaged product's barcode. All three land in the same meal entry shape — name, calories, protein, carbs, fat — so the Dashboard ring doesn't care which one you used.
Photo mode
Take or pick a photo. A vision model names the dish, then that name is matched against the local food database with the same fuzzy matcher used for typed entries — so a photo match and a hand-typed one score identically.
Text mode
Describe the meal in plain language — the AI estimates calories and macros directly.
Barcode
Scan a packaged product. Nutrition comes back from Open Food Facts, including its Nutri-Score (nutritional quality) and Green-Score (environmental impact) letter grades, A through E.
Photo mode can run entirely on-device using a compact vision-language model, so a meal scan doesn't have to leave the phone or wait on a network round trip. It identifies the dish; the calorie and macro numbers still come from the same curated food database — see Nutrition Database — the model's only job is naming what's on the plate.
Save a meal you eat often as a named entry with its own fixed macros, so it's one tap next time instead of a re-scan.
Group several foods eaten together — a plate, not just an ingredient — and log the whole combo as one entry.
Every entry, however it was logged, feeds Meal history — the running record Progress and the AI Coach both read from when they talk about your nutrition averages.