Create AI Images with Mixup: The Mad Libs-Style Photo App
▼ Summary
– Mixup is a new AI photo editor for iOS from Things, Inc. that uses shareable “recipes” as fill-in-the-blank prompts to generate images from photos, text, or sketches.
– The app is built on Google’s Nano Banana model, which maintains user images convincingly without being creepy, and transforms generative AI into a social, game-like experience.
– Users can publish their created prompts and resulting images to a public feed, allowing others to reuse the recipes with their own content and see both the input and output for predictable results.
– Mixup includes a “mixables” feature where users can upload their photos, enabling people they follow to create AI images using their likeness, fostering collaboration among friends or creators.
– The app is free to start with 100 credits, uses OpenAI and Google’s controls for content moderation, and launches globally on November 20 with an invite requirement for access.
A new iOS application named Mixup is transforming how people interact with artificial intelligence for image creation. Developed by the team behind the 3D design tool Rooms, this app introduces a novel “recipe” system that functions much like a Mad Libs game for photos. Users can input their own images, text, or simple sketches, and the app generates entirely new visuals based on customizable, fill-in-the-blank prompts.
Imagine converting a rough doodle into an elaborate Renaissance-style artwork or seeing your cat dressed up in a hilarious Halloween outfit. You could even upload a selfie to experiment with a completely new hairstyle or playfully reimagine a friend in an unexpected scenario. The app builds on Google’s Nano Banana technology, which excels at incorporating user-provided images into new creations without producing unsettling or distorted results. Jason Toff, founder and CEO of Things, Inc., highlighted this capability, noting his team’s experience from previous roles at major tech firms like Google, Meta, and Twitter.
What truly sets Mixup apart is the shareability of its recipes. Instead of starting from scratch with an empty text box, users can browse a public feed of images and the prompts that generated them. Tapping a “Try recipe” button lets anyone reuse those prompts with their own personal photos or drawings. This approach tackles a common frustration with generative AI, the lack of predictability and control over outputs. Toff described this issue as a “slot machine problem,” where each request yields a random result. By displaying both the original input and the final image, Mixup gives users a clearer expectation of what their own creation might look like. A toggle feature also allows side-by-side comparison of the before and after versions, provided the original creator enables this option.
Users can upload their own photos to incorporate into AI-generated images. If you choose to do this, anyone you follow within the app can also create images using your likeness, thanks to a feature called “mixables.” This is designed to encourage friend groups to connect and collaborate, though it could also lead to a community of creators who enjoy seeing themselves reinterpreted in creative, and sometimes bizarre, ways. Privacy remains under the user’s control, simply avoid uploading personal photos or following others if you prefer to keep your image private.
To address concerns about inappropriate content, Mixup employs OpenAI’s moderation tools in addition to the built-in safety controls within Google’s image model. These measures work together to filter out sexual content and violent imagery. The app is optimized for iOS 26 but remains compatible with devices running iOS 18 and newer. A web version or Android release could follow if the app gains sufficient popularity.
At launch, every user receives 100 free credits, valued at approximately $4. Each generated image costs nearly 4 cents to produce. Once the initial credits are used, several subscription plans are available, offering 100, 250, or 500 credits per month. Mixup becomes available for download globally on November 20, though access will initially be invite-only. The app is currently open for preorder on the App Store.
(Source: TechCrunch)
