NutriLens
HomePricingTechnologyPrivacy
Launch scanner
NutriLens

Product

  • Scanner
  • Dashboard
  • Technology
  • Onboarding

Trust

  • Privacy
  • Accuracy
  • Data sources
  • Export

Company

  • Manifesto
  • hello@nutrilens.app

© 2026 NutriLens Labs

Not medical advice — see a registered dietitian for clinical needs

All systems nominal

THE LAB

How NutriLenssees food.

Vision models, an OCR engine tuned for nutrition panels, and a verified data backbone — engineered to show you confidence instead of hiding it.

SCROLL

LABEL OCR · THE DIFFERENTIATOR

Built to read the panel, not guess the plate.

Nutrition panels are rigidly structured — perfect for OCR. Our engine detects the panel, corrects perspective, reads every line, then parses values into 14 nutrient fields with per-field confidence. You always see the parse next to the photo before anything is saved.

  • 14+ FIELDS
  • ~1.2s MEDIAN READ
  • 96% FIELD ACCURACY
Try it in the scanner

Nutrition facts panel being analyzed by the OCR engine

PROVENANCE

Three sources. Zero invention.

Verified nutritional reference

Standard reference entries for calories, macros, and key micronutrients reconciled against verified nutritional food datasets.

VERIFIED DATA

Open Food Facts

Packaged products via Open Food Facts. We look up the barcode code number to return the manufacturer nutritional panel.

OPEN FOOD FACTS

Your confirmed records

Confirmed meal entries are stored securely in your isolated account database. You have full ownership to edit, export, or delete your records anytime.

SECURE LEDGER

Every number wears one of these

  • USDA VERIFIEDstaff-reviewed reference entries
  • LABEL OCRread from the printed panel
  • AI ESTIMATEmodel estimate + confidence
  • OPEN FOOD FACTSbarcode-matched product data
  • MANUALentered by you

HONEST NUMBERS

We publish our error bars.

Photo-based estimation has real limits — plating, sauces, and hidden fats all add error. Most apps hide this. We benchmark on open plate datasets, publish the ranges, and surface low-confidence results as ranges in the UI.

  • Benchmarked in the spirit of Nutrition5k-style public plate datasets
  • Label OCR bypasses estimation entirely — 96% field accuracy
Try the meal scanner

Plate-photo calorie error by dish type

Simple plated meals

Bowls & salads

Mixed dishes

Sauces & hidden fats

shown as a range in the app

  1. 01CAPTURE
  2. 02PROCESS IN MEMORY
  3. 03KEEP THE NUMBERS
  4. 04DISCARD THE PHOTO

The photo is a means. The numbers are the product.

Images are processed in memory and discarded in seconds. No training on your photos without explicit opt-in. Full story on the privacy page.

Read the privacy promise →

MANIFESTO

Nutrition data should behave like evidence.

Every calorie carries a source. Every estimate carries a confidence. Every photo is disposable. We built NutriLens because the category leader ships ads, the AI newcomers hide their error bars, and nobody reads the label. We read the label.

NutriLens is a general-wellness tool. It is not medical nutrition therapy and cannot detect allergens. Consult a registered dietitian or physician for clinical conditions, and always check packaging for allergens.