PhotoWeigh: Estimating Weight From a Photo With AI

The idea: a scale in your pocket

There are plenty of moments when you want to know roughly how much something weighs and there is no scale nearby: portioning food, following a recipe abroad, checking whether a parcel is under a postage limit. PhotoWeigh is one of our own products, built to give a quick estimate from a single photo.

It is listed on the App Store as "Scale for Grams", which is how people describe the need when they search for it.

What the app does

• Estimates the weight of an everyday object from one photo.
• Uses AI visual recognition to identify what the object is.
• Uses a reference item in the frame to establish real-world scale.
• Covers common food and household categories.
• Converts between grams, ounces and pounds on the spot.
• Keeps recent measurements to hand for quick re-checks.

Can an app estimate weight from a photo?

It can estimate it - and the word estimate matters. A photo contains no direct information about weight, so the app has to infer it from two things it can work out: what the object is, and how big it is.

Step 1: what is it?

Two objects of the same size can weigh very different amounts depending on what they are made of - a loaf of bread and a block of cheese, for example. AI visual recognition identifies the object and places it in a food or household category, so the estimate is based on what it is, not only on how big it looks.

Step 2: how big is it?

A single photo has no sense of scale: the same apple looks large up close and small from a distance. Placing a reference item of known size in the frame gives the image a ruler, so the object's real dimensions can be worked out.

Step 3: the estimate

With a category and a size, the app produces an estimated weight and shows it in the user's preferred unit, with grams, ounces and pounds a tap apart.

Setting honest expectations

The most important product decision in an app like this is how it talks about its own output. PhotoWeigh presents every result as an estimate, explains the reference-item step at the moment of capture rather than in a help page, and is positioned for the times a kitchen scale is not at hand - not as a replacement for one when precision matters.

Stack and approach

The app is built in React Native with TypeScript and NativeWind. The capture flow, results, unit conversion, history and paywall share one codebase, so the four-person team could spend its time on the estimation and the capture experience rather than on platform plumbing.

Why we default to one codebase, and when we would not, is in React Native vs native.

Monetisation and reach

PhotoWeigh is free to download, with a one-off Premium upgrade or a yearly plan. It is localised into 41 languages, because weighing food and post is an everyday need everywhere, and units and habits vary by country.

What we would do differently

We would show a range rather than a single number from the first release. An estimate presented as "about 180-220 g" sets a more honest expectation than "200 g", and users trust a tool more when it is upfront about its margin.

Building an AI or computer-vision app?

Photo-to-answer products - estimating, identifying, measuring - are now practical to build on mobile. The difference between a gimmick and a useful tool is mostly in the capture guidance and in how honestly the result is presented.

Validate an idea like this quickly with a 30-day MVP.