Guess What? is a quiz app that uses the u2net Core ML model to detect the subject of a photo and mask it with a red silhouette. Tap Show Answer and the silhouette slowly fades away over three seconds to reveal what was hiding. Five categories — Animals, Food, Vehicles, Scenery, and Tools — are built in, and you can quiz yourself with photos from your own library.
The u2net Core ML model detects the subject of each photo and overlays it with a red silhouette mask. The model loads in the background at launch and the quiz starts automatically when it is ready. All processing is done on-device — no internet connection required.
Tap Show Answer and the silhouette fades out over three seconds using a SwiftUI animation, gradually revealing what was hidden underneath. Take your time guessing before you tap — or tap right away if you give up.
Animals, Food, Vehicles, Scenery, and Tools — over 100 questions built in. Use the left and right arrow buttons to move between questions. After the last question it loops back to the first.
Tap Select Photo to choose any image from your photo library. The AI masks the subject of your chosen photo and turns it into a quiz question. Photo library access permission is required.
At launch, the u2net Core ML model is loaded on a background thread. Once ready, a VNCoreMLRequest processes each photo — the resulting MLMultiArray mask is converted to a grayscale UIImage and then composited as a red silhouette using CIFilter blendWithMask. The whole pipeline runs locally with no server calls.
The mask image is layered over the original using a ZStack. Its opacity is driven by a @State variable that starts at 0.0. Tapping Show Answer triggers a withAnimation(.easeInOut(duration: 3.0)) block that animates the opacity to 1.0, causing the silhouette to fade out gradually and expose the hidden subject.
Photos for Animals, Food, Vehicles, Scenery, and Tools are bundled as app assets. The left and right arrow buttons cycle through them in order. When you reach the last question the next tap loops back to the first, so you can keep playing without hitting a dead end.
Tapping Select Photo opens a PHPickerViewController limited to single image selection. The chosen UIImage is passed directly to the same prediction pipeline used for the built-in questions. The silhouette resets and the AI generates a fresh mask for your photo.
The AI model loads in the background at launch. When it is ready the first quiz question appears automatically. Use the arrow buttons to browse questions or tap Select Photo to use your own image.
Look at the red silhouette and think about what might be hiding underneath. The shape and size of the mask are your only clues.
Tap Show Answer and watch the silhouette fade away over three seconds. Once you see the answer, tap an arrow button to move to the next question — the silhouette resets automatically.
"My kids love this. We take turns guessing and it always turns into a fun argument about what the shape could be. Using family photos as questions is the best part — the AI makes a silhouette of whatever you pick and it actually works really well."
"Simple concept executed well. The food category is surprisingly hard — some silhouettes are really tricky. The slow reveal animation is a nice touch. Good for a few minutes of entertainment whenever you have downtime."
"Tried it with my own landscape and pet photos and the silhouettes came out clean. The app is straightforward — no menus to dig through, just the quiz. Works exactly as described."
Find answers to common questions below.