Apple's Updated Genmoji: Mixing Real Emojis to Create Custom Ones

Apple has given its Genmoji feature a significant upgrade. Genmoji — custom emoji that you generate by typing a description — can now work with existing emoji as starting points. Instead of describing an emoji from scratch, you can now pick one or more real emoji, blend them together, and add text instructions to create something entirely new.

What's New

When Apple first introduced Genmoji in June 2024 as part of Apple Intelligence, the process was simple: you typed what you wanted, and the system created an emoji-style image for you. The update announced in September 2025 adds a new layer to this. As Apple detailed in their announcement, you can now use existing emoji as visual ingredients. You select one or more emoji, combine them, add a text description, and the system generates a fresh Genmoji that blends the look and meaning of your source emoji with your written instructions.

Think of it like collage-making: instead of describing a picture to someone and hoping they draw what you meant, you can show them some reference images, point to elements you like, and then describe what you want them to change or combine.

Where You'll Use Them

Genmoji work wherever you'd expect them to — directly in text messages, as stickers, or as Tapback reactions (those quick emoji responses you get when you long-press a message). This hasn't changed in the update. What's new is how you create them. The system lets you insert a freshly generated Genmoji straight into a message thread without needing to save it to your camera roll first. That removes a step that used to get in the way of quick, spontaneous expression. Tapback reactions, which Apple has gradually expanded beyond the original six fixed emoji, now accept Genmoji too — so your reaction can be fully custom.

How the Technology Works

The technical picture here is worth understanding. To make this work, Apple's system needs to accept two kinds of input at once: text descriptions and visual references (the emoji you picked). This is called "image-conditioned generation" — the model receives both a description and a visual reference, and it uses both to shape what it creates. How much weight the source emoji carry — whether they tightly constrain the style, or loosely influence just the general direction — isn't spelled out in Apple's public statements, but the language they use ("mix together," "combine with descriptions") suggests the emoji matter more than just tagging metadata.

This matters if you're following how Apple builds AI into its products. It means the system running Genmoji likely includes a more sophisticated type of AI model called a multimodal encoder — one that understands both images and text simultaneously. Whether this runs entirely on your device (Apple Silicon chips like the A18 Pro are powerful enough), or uses Apple's Private Cloud Compute servers while protecting your privacy, remains unclear. Apple hasn't said.

A Familiar Pattern

There's a pattern in how consumer software adopts generative AI that's worth recognizing. Tools typically start with text-only prompts because that's the quickest way to ship something safe. Then, once the team is confident the outputs are good and guardrails are solid, visual inputs arrive in version two or three. Text-to-image generators followed this exact path: early versions took text prompts, later ones took reference images, then style images, then compositional guides. Apple is following the same progression, but with a controlled vocabulary. Instead of accepting any image, Genmoji only draws from the existing Unicode emoji set — a fixed, licensed collection. That constraint likely made it safe to launch the combinatorial feature now, before Apple opens things up to more flexible image inputs.

Who Can Use This

Apple Intelligence requires fairly recent hardware: an iPhone 15 Pro or any iPhone 16, an iPad with an M1 chip or newer, or a Mac with Apple Silicon. These are the devices with enough processing power to run the AI models needed for near-instant or quick generation. The September 2025 update applies to the broader Apple Intelligence rollout, and Apple hasn't specified whether combinatorial Genmoji require the same hardware floor, but that's the safe assumption.

What This Means

For everyday users, this is straightforward: a faster, easier way to create personalized emoji for messages and reactions. For anyone building software that runs on Apple platforms, the larger signal is that Apple continues to expand what you can generate and customize, but it's all happening inside Apple's own walls. The generation tools are Apple's, the outputs go through Apple's systems, and they live in Apple's messaging and reaction features. Apple has deliberately kept it that way, and that choice determines what's possible for third-party developers.

Looking at the bigger picture: the Unicode Consortium — the organization that standardizes emoji across the industry — releases new emoji through a formal, slow process to ensure interoperability. Genmoji runs in parallel to that. One creates portable, universal emoji through consensus; the other creates personal, instant emoji on demand. Both can coexist, and that shift in how we create visual language on phones is a meaningful one. The next question is whether Apple will eventually bring these tools to other messaging contexts beyond its own apps, or keep them purely within its ecosystem.