Snap a photo. Build the habit.
Panotxa turns a photo of your plate into a meal-quality score, then coaches your trends — not your guilt. No barcode scanning, no weighing, no obsessive calorie math.
How it works
Photograph your meal
One photo per dish. Panotxa works with real plates — home cooking, restaurants, even a shared paella for the table.
AI reads the plate
Vision AI extracts objective signals — vegetables, protein quality, whole grains, processing level, portion, drinks — not just a calorie guess.
Trends do the coaching
Each meal gets a transparent 0–100 quality score. Your 7-day Momentum and 42-day Habit show where your eating is really going.
The science: training load, applied to eating
Panotxa borrows the math endurance athletes use to balance training stress and fitness — exponentially-weighted moving averages — and applies it to meal quality. The goal is to coach trends, not to make you obsess over a single plate.
| Training concept | Panotxa metric | Meaning |
|---|---|---|
| Individual effort | DQ · Dish Quality | 0–100 per dish, from one photo |
| Training stress (TSS) | MQ · Meal Quality | 0–100 per finished meal, all dishes combined |
| Acute load (ATL) | Momentum | 7-day trend of your meal quality |
| Chronic load (CTL) | Habit | 42-day baseline — who you are lately |
| Stress balance (TSB) | Wiggle Room | Momentum minus Habit — earned flexibility |
90 example days through the app's real math
Every score maps to a plain-language zone
Seven transparent dimensions, one score
No black box: every dish is scored on published formulas over the extracted signals. The AI's own opinion is blended in at just 30% — the transparent math stays in charge.
Fair to real life
- Snacks are judged as snacks — an apple scores ~85, a plain coffee ~90. Logging a good light choice is never a punishment.
- Shared-table food (tapas, a paella for four) is attributed as one personal portion, not the whole table.
- Protein, calorie and weight-trend targets follow published sports-science positions (ISSN, ESPEN) and the Harvard plate — with sources.