# Meal Photo Macros

> Photo of a meal -> each food named and matched to the USDA FoodData Central table, a portion RANGE in grams per item, and calorie, protein, carb, fat and fibre ranges computed by the page from the table. Editable portions, a daily log, CSV and Markdown export. An estimate, not medical advice.

URL: https://meal-photo-macros.skillsafe.ai/
API: https://meal-photo-macros.skillsafe.ai/api.html

## What it does
- One paid step: an AI model reads the photo, names each food, picks the closest table id (or gives its own per-100 g values for a dish the table lacks, checked against the Atwater factors), sizes a low-high portion in grams, and lists likely hidden extras such as cooking oil.
- The page does all the arithmetic: low grams x per-100 g -> low, high -> high. Totals are ranges, rounded so they never look more precise than a photo allows.
- Free: four worked examples (including a no-food photo and a re-check with a correction note), adding foods by hand from the table, editing any portion, the daily log.
- Re-check: write what the photo cannot show ("diet cola", "two eggs") and run again; the model keeps what the note does not dispute.
- Daily log: per signed-in account (guests: this browser only), with an optional energy target shown as a range against it.

## Pricing
No API key needed - pay per run with SkillSafe credits. The estimate is shown as reserved before the run; the charge after is what the run used.

## Data
Nutrient values per 100 g: USDA FoodData Central, SR Legacy (April 2018), public domain. 319 foods. Photos are shrunk in the browser and location data is removed before upload.

## Output contract
JSON: status (ok | no_food | unclear), meal_title, scale_cue, photo_notes, items[] {name, food_id, grams_low, grams_high, confidence, seen, hidden, per100}, questions[].