Barcode scanning is a slow, miserable interface that doesn't work for anything you didn't buy in a package. Restaurant food, leftovers, home cooking, fruit — all useless. CalBurndown skips the scanner entirely. Three faster ways:
1. Type it in plain English
Tap + → Log meal → type whatever you ate:
two scrambled eggs, slice of sourdough, black coffee
The AI parses it into items, looks each one up in a 1.4M-row open food database (Open Food Facts + USDA), and gives you a calorie estimate with macros. You can edit each line before saving.
When it's accurate:
- Common foods with standard portions ("medium banana", "8 oz chicken breast")
- Restaurant chains by name ("Chipotle chicken bowl with brown rice")
- Specific brand names ("Trader Joe's tikka masala")
When to be skeptical:
- "A sandwich" — too vague
- Mixed plates with no portion sizes — better to use the photo path
2. Say it out loud
Same parser, voice input. Tap + → Log meal → microphone icon. Useful when:
- You're driving (don't actually drive and type)
- You're cooking and have wet hands
- You ate enough things that typing is annoying
Voice transcription happens on-device when possible (Chrome/Safari Web Speech API) — your audio doesn't leave your phone in those cases. The transcribed text then goes to our AI parser, same as typing.
3. Photograph it
Tap + → Log meal → camera icon. Snap a normal photo of your plate. The AI:
- Identifies each visible food
- Estimates portion sizes (more accurate with a hand calibration)
- Returns calories + macros per item
When the photo path wins:
- Restaurant plates where you don't know the recipe
- Home cooking with mystery ingredients
- Anything where you'd otherwise just guess
EXIF and GPS are stripped before upload. We don't see your kitchen or your location, only the cropped food image.
Which method should you use?
| Situation | Best method |
|---|---|
| Quick snack you know cold | Type |
| Restaurant entree | Photo |
| Driving | Voice |
| Mixed plate at a friend's | Photo |
| Brand-name packaged food | Type |
| A handful of nuts | Type |
Accuracy expectations
Aim for ±15% per meal. Over a week, errors largely cancel out — what matters for fat loss is consistent direction, not perfect daily numbers. If you're routinely off by more than 25%, it's almost always portion-size error, not the calorie database.
That's why hand-based portion estimation matters more than tracking to the gram.