Photo your plate, log in 2 seconds: how Volya killed manual food entry
MyFitnessPal makes you tap 12 times to log dinner. We made it one photo. Here's how.
Logging food is the chore everyone abandons. Average dropout for food trackers is week 3 — and the #1 reason is friction: search, scroll, pick, edit grams, save, repeat for each item.
The Volya approach
- Open the menu page.
- Tap 📷 Snap plate.
- Take a picture of what's in front of you.
- Edit grams if the AI guessed wrong (auto-rescales macros).
- Hit Log.
That's it. 2–4 seconds end to end for a typical plate. Eva (our AI coach, powered by Gemini 2.5 Flash under the hood) returns per-item grams + protein/carb/fat + meal-type guess. We don't store the photo — only the structured numbers.
Why it works
The hard part isn't recognising "chicken" vs "tofu" — that's solved. The hard part is grams. We trained the prompt to estimate portion sizes against common plate references (10-inch plate ≈ 25 cm, fist ≈ 200 g chicken, palm ≈ 100 g rice).
When it fails
- Soups + smoothies: the camera sees liquid, not ingredients. Use search for those.
- Mixed bowls (poke, buddha bowls): items overlap; expect 15–20% grams error.
- Dark restaurants: lighting kills accuracy.
For 80% of meals (standard plates at home or at work), it's more accurate than your manual estimate.
What it costs
Nothing. Eva's free tier (Gemini 2.5 Flash backend) handles ~3 plates per minute per user. No credit card required.