iOS 18.2 brings a new app called Image Playground that lets you create often whimsical images using AI. Apple has also provided image generation tools in iOS, including the ability to create new emojis in messages.

With Image Playground you can easily create (quirky) AI images. The new app comes pre-installed with iOS 18.2 if you have an iPhone 15 Pro or Pro Max or an iPhone 16 model. When you open the app for the first time, you will be asked to set up image creation with Image Playground, Genmoji, and Image Wand.

Tap the plus button at the bottom of the app and enter a description of the image you want. Image Playground will then create an image based on your keywords. Some variations will be generated and you can swipe left and right to see what came of it. You can add more criteria to your initial image request in the text box below or choose from a few suggestions.

The results are, frankly, ridiculous. Senior editor Caitlin McGarry experimented and asked to create an image of “Caitlin drinking a strawberry matcha.” The result was a cartoon character version of Caitlin with a strawberry growing out of her lip and matcha dripping down her chin.

After some experimentation, we found that Image Playground works best for creating images of people. You can do this by selecting a real person from photos or using the Appearance option to select a skin tone and visual style.

However, Image Playground is not intuitive to use. When Image Playground generated anything other than humans, it was significantly less consistent in our testing. Requesting pictures of octopuses produced a handful of cute results and a dozen scares. When I asked about a dog with coffee, I got several images of a cartoon dog with a cup in his hand, but also even more frightening images of the dog’s face splitting in half and a stream of coffee pouring into a waiting cup .

And it’s not always clear when you create a new image or modify an existing one. Sometimes generating an image just doesn’t work. Image Playground, like all Apple Intelligence features, is currently in beta, so you may find occasional use strange or frustrating.

Creating AI-generated emojis is extremely fun. Tap the emoji button in the bottom left corner of the iOS keyboard. A new search box will appear at the top with the telltale glow of Apple Intelligence. Enter an existing emoji and it will simply search for it. However, when you describe a new emoji, iOS creates what Apple calls a Genmoji.

Comparing the results to Image Playground’s results, we found fewer botched details and less nightmare fuel with Genmoji. Simple queries work best (“submarine” versus “Max Eddy in a submarine”), and mashups will likely take a few iterations to achieve the best results, but given the many custom emojis in Wirecutter’s office Slack we believe this feature could be a huge success.

The Image Wand tool turns sketches into AI-generated images. The Image Wand looks like a magic wand and appears in the Drawing Tools palette in places where you can draw, such as the Notes app. We tried this out by doodling an arched structure, selecting the magic wand, and then circling our drawing. The tool then asked us to describe in words what we were looking for. In a few seconds, iOS presented some options to replace our drawing. If there is text near your drawing, Apple will suggest those words as changes to the AI-generated image; This can be useful if you leave important additional context for the image out of the image prompt. You can use the image stick to circle blank spaces to easily create an image from a prompt, or circle already written text to create an image.

However, sometimes parts of our requests were ignored – for example, a structure was created with any number of columns other than the three we requested – but some of this could be due to our poor doodling skills. It can be interesting to see the results, but for school or business projects you’ll probably want to stick with professional clipart.

Apple is putting some surprising limits on AI image generation. As we expected, requests for offensive images or the use of insults are not permitted. However, we were surprised to find that we couldn’t produce an image of two men kissing, or of a man and a woman kissing. In fact, we could only create one image of one person at a time. However, images of multiple cats, submarines, etc. can be created.

Even more surprising was that words that might have been understood negatively were also excluded from our queries. We couldn’t create images of people looking sad, angry, or crying. No matter what we asked, everyone Apple Intelligence came up with either smiled or had a blank, neutral look. Requests to make a “fat,” “tall,” “large-bodied,” or “plus-sized” person were rejected or ignored. It is impossible to say whether all of these limitations are intentional or just idiosyncrasies of the imaging models. But it feels like Apple’s efforts to just create innocuous images produce results that are generally boring. Even Apple’s custom Memoji allowed you to create angry or sad faces.