Mixup: AI Image Generator - Photos, Text & Doodles

Introducing Mixup: A Novel AI Photo Editor

The team formerly at Google, responsible for the 3D design application Rooms from Things, Inc., has unveiled its newest creation: Mixup, an engaging AI photo editor. This iOS-exclusive application empowers users to generate innovative AI-driven images through the utilization of “recipes,” functioning as fill-in-the-blank prompts for photos, texts, or sketches.

Creative Possibilities with AI Recipes

Users can leverage Mixup to transform a simple doodle into a stunning Renaissance-style artwork, or reimagine their beloved pets in amusing Halloween attire. Selfies can be altered to explore different hairstyles, or more whimsical transformations – such as converting a friend into a representation of Italian internet culture – can be realized.

Built on Google’s Nano Banana

The application’s foundation is built upon Google’s Nano Banana model. However, the unique “recipe” format introduces a fresh method of interacting with the model, effectively converting a generative tool into a collaborative, interactive experience.

The Power of Maintainable Image Generation

Jason Toff, founder and CEO of Things, Inc., highlighted Nano Banana’s key strength: “The capability to integrate your image while maintaining a convincing result, avoiding an unsettling effect, was groundbreaking.” Toff’s experience includes roles at prominent tech companies like Google, Meta, and Twitter, where he managed product development.

Shareable AI Prompts

A particularly appealing aspect of Mixup is the shareability of its “recipes” – the AI prompts created by users. This fosters a community-driven approach to AI image generation.

Addressing the Creative Block in AI

Toff observed a common challenge with current AI tools: “Generative AI is incredibly potent, yet often presents users with a blank text box and the expectation to devise something creative. What does one even write?”

He explained that Mixup circumvents this issue by allowing users to simply fill in the blanks of pre-existing, successful prompts, rather than starting from scratch.

A Public Feed for Inspiration and Reuse

Users have the option to publish their created prompts alongside the resulting images to a public feed, or to download them for private use. Other users can then access the feed, view a photo, and utilize the “Try recipe” button to generate their own image using the same prompt and their own content – be it a photo, text, or a doodle created with the app’s built-in drawing tool.

Controlling the Unpredictability of AI

The team believes that displaying the photo alongside its corresponding recipe can mitigate the often-unpredictable nature of generative AI images.

Combating the “Slot Machine” Effect

Toff described a common frustration with generative AI as a “slot machine” problem: “You press a button, get a result, press it again, get something different, and feel a lack of control over the output.”

Mixup addresses this by presenting both the image and the prompt in a unified view, providing users with a clearer expectation of the potential outcome. Creators can also enable a toggle to display both the original and generated images for comparison.

“Mixables” and User Likenesses

Similar to OpenAI’s Sora, Mixup allows users to upload their own photos for use in AI-generated images. This feature, termed “mixables,” enables users who follow each other to create images featuring each other’s likenesses.

The company anticipates that friend groups will utilize this feature, but also acknowledges the potential for a creator class to emerge, willing to have their image creatively repurposed.

Content Moderation and Safety

Mixup incorporates OpenAI technology for content moderation. Additionally, it relies heavily on Google’s built-in controls within its image model to prevent the generation of inappropriate content, such as sexually explicit material or violent imagery.

Availability and Pricing

Mixup is currently optimized for iOS 26 and compatible with iOS 18 and later. Future versions for web or Android are possible depending on its success.

New users receive 100 credits, equivalent to $4, with each image generation costing approximately 4 cents. Subscription tiers offering 100, 250, or 500 credits per month are available once the initial credits are depleted.

Launch Details and Early Access

The application will launch on the App Store globally at midnight on November 21, but access will initially require an invitation. TechCrunch readers can use the code TCHCRH (while supplies last) to gain early access. Mixup is currently available for preorder.

Note: This article was updated following publication to reflect the correct launch date of November 21.