If you often eat Chinese food, do not choose a food-logging app based only on whether it recognizes “kung pao chicken.” What matters in daily use is whether it can handle mixed dishes, let you correct portions, and make cafeteria meals, takeout, and shared dining easy to log. Its nutrition data should also include relevant Chinese food sources. nosh can help you capture a meal quickly, but no app proves it works for Chinese food simply by saying so. Test it with the meals you actually eat.
A plate of grilled salmon often has clear boundaries: fish, vegetables, and a staple each sit in their own area. Chinese meals are rarely that simple. A photo cannot separate how much oil the egg absorbed in tomato and egg stir-fry, the ratio of pork to vegetables in fish-fragrant shredded pork, or how much sauce from a rice bowl soaked into the rice.
So supporting Chinese food is not a dish-name quiz. Recognizing the right name only shows that the app broadly understands what the dish is. It does not mean the app knows how this serving was cooked, how much you took, or how much you finished.
The same dish name can describe very different meals
Braised pork made at home may be leaner and less sweet, while a restaurant version may contain more oil and sauce. One person may stir-fry greens quickly with a small amount of oil, while another uses much more. The name can stay the same while the ingredient ratio, preparation, and portion all change.
That is why counting how many Chinese dishes an app recognizes is not enough. A large food database cannot reveal cooking details that are absent from the photo. When information is uncertain, a reliable app should let you correct what you know instead of presenting an unexplained, overly specific answer.
nosh identifies food from photos and estimates portions, calories, and nutrition information. Its nutrition data mainly refers to sources including the Chinese Food Composition Tables, USDA FoodData Central, and the FDA. Chinese food data is an important foundation, but it does not mean every family or restaurant recipe can be reconstructed accurately.
See how it handles a meal with everything mixed together
The most common unit in a Chinese meal is not one standardized plate. It is a combination: one meat and two vegetables on a cafeteria tray, rice pressed against dishes in a takeout box, toppings piled over noodles, or food shared from the same hot pot.
During a trial, look at three details: whether the main foods are missing, whether overlapping foods are collapsed into one dish, and whether you can adjust the amount you personally ate. Recognizing an entire hot-pot table does not mean the app knows how many slices of beef, pieces of tofu, or bowls of dipping sauce you had. For shared meals, logging the main foods you actually ate is often more useful than trying to calculate the whole table precisely.
A photo records how the food looked when it arrived, not necessarily what you finished. Leaving one third of the rice, drinking only a few sips of soup, or removing the coating from fried food usually requires your own correction. nosh currently lets you edit food names, calories, portions, and log times. Correct the items that change the overall picture first; you do not need to reconstruct every bite in grams.
Do not forget breakfast, snacks, and drinks
Many people test Chinese food recognition with large main dishes. The items more likely to disappear from a log are soy milk, rice rolls, steamed buns, an afternoon milk tea, or a snack eaten while walking. They may not be difficult to recognize, but they are easy to miss when no photo was taken.
In addition to the camera, check whether the app can import from your camera roll and accept a quick text or voice entry. nosh supports all of these. If you remember only after eating, “two pork buns and one unsweetened soy milk” is usually more useful than repeatedly searching for every ingredient in the filling.
Test the app with five meals from your own life
You do not need standard reference images. Choose five meals you are likely to eat soon: one home-cooked stir-fry, one cafeteria plate, one rice bowl or noodle dish, one shared meal, and one breakfast or afternoon snack.
Do not record only whether recognition was right or wrong. Also ask:
- When a main food is wrong, can you quickly correct the name and portion?
- When oil and sauces are invisible, does the app present an estimate as if it were a fact?
- If you forget to take a photo, can you add the meal easily with text or voice?
- When you look back the next day, can you still understand what you actually ate?
If an app recognizes neatly plated single items but makes you re-enter everything on a cafeteria tray, it may not fit your life. On the other hand, an approximate number can still be useful when the main foods are present, corrections are easy, and logging feels light enough to continue.
For people who often eat Chinese food, the standard can be simple: the app does not need to pretend it understands every spoonful of oil. It should understand that Chinese dishes overlap, recipes change, and portions belong to the person eating. nosh is not trying to force a whole table into one exact equation. It helps you save the meal first, then gives the parts you know back to you to correct.