How does Waffly estimate?

From capture to estimate

You capture a meal — by voice, text, barcode or photo. Waffly normalizes your input into a food with a quantity and estimates calories and macros. Waffly compares that estimate against general reference values based on DGE reference values.

Accuracy & limits

Values are reference estimates for adults — not individual nutrition advice and not medical advice. When uncertain, you choose an estimation mode in settings that influences how Waffly estimates portion sizes.

Data flow

Your voice input is converted to text on your device — the audio never leaves it. For estimation only the text (up to 4,000 characters) is sent to the backend. Photos are stored in an EU data center and sent to Google Gemini for estimation. Barcode scans go directly through the open Open Food Facts database. For anonymous product analytics Waffly uses PostHog on EU servers; you can disable this in settings.

Providers

Waffly uses Google Gemini as its primary AI provider, with OpenAI as a fallback in exceptional cases. With Bring your own key, your own OpenAI-compatible provider processes requests directly from your device — your key stays on your device.