How to track calories while traveling internationally (when half the food has no English label)

Tracking calories in your home country is hard. Tracking them in a country where you can't read the menu, the portion conventions are different, and the local cuisine has approximately zero entries in the USDA database is a different problem entirely.

This is the situation photo-based AI tracking was designed for. Here's how to use it.

Why barcode and text-search trackers fall apart abroad

The legacy trackers (MyFitnessPal, Lose It!, Cronometer) all rely on:

• Barcodes (mostly US/UK products in the database)
• Brand names (locale-specific)
• Restaurant chains (US-centric)
• Standard portion sizes (American conventions)

In Tokyo, Bangkok, Lima, or Marrakech, all four assumptions fail. The food is unbranded, cooked in front of you, served in unfamiliar portions, and named in a language the database doesn't index.

You can spend 20 minutes per meal trying to reverse-engineer it. Or you can take a photo.

The photo-tracking advantage abroad

Photo-based AI looks at the food, not the label. A bowl of pho is a bowl of pho whether it's named in Vietnamese or English. The vision model identifies the noodles, broth, herbs, and protein, estimates portions from depth and visual size, and gives you a number.

The number won't be perfect. But it will be in the right ballpark — usually within 20% — which is better than the alternative (skipping the log entirely).

The framework: snap, adjust up, log

For international meals:

1. Snap the dish. Top-down works best.
2. Read the AI's interpretation. It might say "rice noodle soup with beef and vegetables." Close enough.
3. Adjust upward by 10–20%. Local cuisine often uses more oil, butter, or sugar than American cooking. Bias the estimate up.
4. Log it. Move on.

Don't hunt for the "correct" entry. There usually isn't one.

What about street food?

Street food is often the healthiest option when traveling — fresher ingredients, less hidden butter than restaurants. It's also impossible to look up in a database.

Photo log it. Add a note about the meat type (chicken, pork, beef, fish) so the AI's portion estimate is on the right calorie base. Done.

Cuisine-specific calorie traps

Some cuisines are sneakily high-calorie even when they look light:

• Thai: coconut milk-based curries are 500–800 cal per bowl
• Indian: ghee, butter chicken, naan with butter — 1,000+ cal entrees
• Mexican (real): lard in rice, cheese in everything — closer to American Mexican than expected
• French: butter on/in everything; an "innocent" omelet has 50 g butter
• Italian (real): olive oil portions are 2–3 tbsp (300+ cal) on a single salad

And some cuisines are calorically lighter than they look:

• Japanese (traditional): sushi, sashimi, soba — relatively low cal density
• Vietnamese: pho, banh mi, fresh spring rolls — moderate density
• Korean (traditional): rice + protein + many small vegetable banchan — moderate
• Greek (real, not Americanized): lots of vegetables, olive oil, fish — moderate

When in doubt, default to the higher cuisine assumption. Adjust over the trip as you calibrate.

The drink translation problem

International drinks vary wildly in size and sugar:

• European beer: often 500ml, not 12oz — 200 cal not 150
• Italian aperitivi: Aperol spritz, negroni — 180–250 cal
• French wine: often poured larger than US — 200 cal/glass
• Asian milk teas: can be 400+ cal each
• Mexican aguas frescas: 150 cal/cup

Photo log liquids too. You'd be surprised at the AI's accuracy with unfamiliar drinks if the glass is in frame.

The hotel-breakfast issue (international edition)

The "European breakfast" of meats, cheeses, breads, and pastries is calorie-dense:

• Croissant: 270 cal
• Two slices prosciutto + cheese: 200 cal
• Brioche: 250 cal
• Yogurt + honey: 200 cal
• Coffee + foam: 80 cal

A typical European breakfast plate clocks 800+ cal. Plan accordingly, or load the plate with eggs + fruit instead of pastries.

When to abandon precision

If you're on a 7-day trip, your goal should be trend tracking, not precision. The questions that matter:

• Did I eat roughly maintenance, or way over?
• Did I drink way more alcohol than usual?
• Did my move pattern roughly match my eating pattern?

Approximate answers to these three questions are more valuable than precise answers to the question "exactly how many grams of pad thai did I eat?"

The CalorieScan AI international workflow

In the app:

• Set the language preference to your home language
• Use photo-log mode for everything (skip search, skip barcode)
• Add a note in your own language about what the dish is — useful when you review later
• At the end of the trip, scan your weekly summary; trends tell you more than meal-by-meal numbers

The honest summary

International tracking is the use case where AI photo recognition earns its keep. The barcode-and-database approach was never going to work in Hanoi.

Snap, adjust up, log, move on. Be approximate. Trust the trend.

The exact calories of a Bangkok noodle stall don't exist anywhere. The approximate calories exist on your camera roll. Use them.