Calorie databases were built around burgers and salads — which is why logging 滷肉飯, hotpot or a hand-shaken drink feels impossible. Here's the method that actually works: components, hand portions, and honest numbers for the dishes apps guess at.
Search "lu rou fan" in a typical tracking app and you'll find entries from 300 to 900 kcal — user-submitted guesses, all confidently wrong in different directions. Mixed dishes vary by cook, portion and fat. The fix is to stop logging the name and start estimating the components: rice + protein + fat is a far better model than a database entry.
Nobody weighs food at a night market. Hand portions are with you always: a fist of cooked rice ≈ 200 kcal, a palm of meat ≈ 150–200 kcal, a thumb of oil, sauce or peanut butter ≈ 100 kcal, a cupped hand of noodles ≈ 200 kcal. A small 滷肉飯 becomes: one fist of rice (200) + half a palm of braised pork (100) + one to two thumbs of braising fat (100–200) ≈ 450–550 kcal. That's a real number you can use.
A 700ml full-sugar pearl milk tea is 500–650 kcal — nutritionally, a bowl of braised pork rice you drink through a straw. Sugar level and toppings swing it by hundreds of calories, so log the cup like a dish: half sugar saves ~100–150 kcal, no pearls saves ~150–200, and unsweetened fresh tea is nearly free.
Don't try to log the table — log what lands in your bowl. Count your actual pieces (each fish ball ≈ 40–50 kcal, each slice of pork belly ≈ 80–100), add one thumb (~100 kcal) per bowl of sesame or shacha dipping sauce, and remember the broth you sip: clear broths are light, butter-based mala broth is not.
Vision AI can now identify dishes and estimate portions from one photo — the method above, automated. The catch: most scanners were trained on Western plates, so their accuracy collapses exactly where you eat. Whatever tool you use, correct its wrong guesses; a good scanner learns your plates.
Honest ranges, not false precision. Portion and cooking fat are always the swing factors.
| Dish | Typical portion | Estimate |
|---|---|---|
| 滷肉飯 braised pork rice | small bowl | 450–550 kcal |
| 牛肉麵 beef noodle soup | regular bowl | 600–800 kcal |
| 鹽酥雞 popcorn chicken | night-market bag | 500–700 kcal |
| 小籠包 soup dumplings | per basket of 8 | 400–500 kcal |
| 珍珠奶茶 pearl milk tea | 700ml, full sugar | 500–650 kcal |
| Ramen (tonkotsu) | regular bowl | 550–800 kcal |
| Bibimbap | stone bowl | 550–700 kcal |
| Pad thai | street portion | 600–800 kcal |
FoodLens is our lab experiment: photo-in, honest-numbers-out, trained around Asian cuisine — night-market dishes, shared plates, and drinks with sugar and ice levels, logged in 中文 or English. Your signup is the vote that gets it built, and your "impossible to log" dishes become our test set.
Full pitch: FoodLens experiment page
Small bowl roughly 450–550 kcal, large 600–750 — the braising fat over the rice is the swing factor, not the rice itself.
A 700ml full-sugar milk tea with pearls runs 500–650 kcal. Half sugar saves ~100–150, skipping pearls ~150–200, and unsweetened fresh tea is under 100.
General scanners often misread mixed plates, soups and drinks because their training is Western-heavy. Asian-specific training data plus your corrections is what closes the gap — exactly what FoodLens is for.