Milk tea, coffee, and other drinks are difficult to estimate not because the cup is blurry, but because a photo usually shows only its shape and color. It may not reveal the cup size, milk, syrup, foam topping, or add-ins. For drinks, order options, cup labels, and nutrition facts often provide more information than the finished drink itself.

A photo in nosh can first preserve what you drank. Then adjust the entry using the actual size and how much you finished. Do not ask AI to recover an entire recipe from a pale-brown liquid. The size, sweetness, milk, and toppings you chose are the most useful clues to add.

A liquid hides many ingredients behind one color

Two similar-looking light drinks could be an Americano with a splash of milk and a whole-milk latte—or a milk tea with syrup and cheese foam. Solid foods often have visible boundaries. Once sugar, milk, and a drink base are mixed together, a camera cannot reliably separate them again.

Even a clear cup may not help much. Ice volume, a tapered base, and a partially finished drink all affect how full it appears. Without a dependable size reference in a single photo, a taller cup is not necessarily larger, and a full cup does not always contain the same volume.

A drink photo is therefore better at answering “What is this?” than “Exactly how much of every ingredient is in it?”

“Half sugar” does not mean half the calories

Choosing half sugar usually changes only part of the added sugar or syrup. It does not halve the milk, foam, tapioca pearls, coconut jelly, or drink base. Stores may also use different formulas for full, half, and less sugar, so “half sugar” is not a universal nutrition conversion.

Likewise, “no ice” describes the ice level; it does not tell you exactly how much extra drink was added. Oat milk, rich dairy blends, and coconut milk are not interchangeable ingredients either. Recording these words makes the drink identity more accurate, but the name alone cannot produce an exact calorie count.

For milk tea, the most useful details are usually the actual size, sweetness, milk, add-ins, and how much you drank. Stop when the record is useful; you do not need to reverse-engineer the order into a lab recipe.

For packaged drinks, read the label instead of guessing what is already written

For bottled milk, juice, soda, or ready-to-drink coffee, a matching nutrition label is the more direct source. Check whether values are listed per 100 milliliters or per serving, then compare that with the bottle size and how much you actually drank.

If the label lists energy per 100 milliliters and the bottle contains 450 milliliters, the label number is not the total for the bottle. If you drank half, do not record the whole bottle. A different flavor, package size, or product version may also make an old screenshot or similar-looking package a poor match.

For a made-to-order drink without a complete label, use current information for the same product and size when available, then combine it with the photo. The closer the source is to the drink in your hand, the more useful it is.

The cup remains after the drink is gone, but half the evidence may be missing

An empty cup can preserve the brand, size, label, and traces of toppings, but it cannot show the original fill level. Instead of repeatedly photographing the empty cup, recall the order: medium or large, any syrup or add-ins, and whether you finished it, drank half, or took only a few sips.

Coffee is similar. A small creamer added to black coffee, a different milk in a latte, or sugar added halfway through may not appear in the final image. Recording what actually changed is more reliable than treating every brown coffee as the same standard cup.

In nosh, check four things first

After photo recognition, check the drink name, size, add-ins, and the amount you actually drank. nosh currently lets you edit the name, calories, portion, and log time. Without a photo, you can also use text or voice to record something like “medium hot latte, no sugar, finished.”

A nosh photo estimate is useful for capturing a drink quickly, but a clear cup does not reveal every ingredient inside. Let the photo preserve the moment, let the label and order supply known facts, and leave what remains unknown as an estimate.

References

Related reading