Skip to content
IFAI

AI tools

Photoshop gets Firefly AI with Generative Fill

Published Reach 43

Adobe just brought its Firefly generative AI directly into Photoshop through a new feature called Generative Fill. Users can now expand images, add objects, or remove unwanted elements simply using text prompts across separate layers. Because Firefly is trained on Adobe Stock and open-license content, generated assets are designed to be commercially safe. A full rollout is slated for the second half of the year.

Photoshop gets Firefly AI with Generative Fill
Image: theverge.com

Instead of bouncing back and forth between a browser generator and your photo editor to stitch assets together, Adobe is embedding the model right into your layers.

Adobe has integrated its Firefly AI image generator directly into Photoshop with a new feature called Generative Fill. The tool launched in beta today and is scheduled for a full release in Photoshop in the second half of 2023.

Prompt to expand, right on your layers

Generative Fill operates just like any regular Photoshop tool, working inside individual layers of your project.

Users can extend the borders of an existing image through outpainting, or add and remove specific elements simply by typing text prompts. Every time you generate, the tool serves up three distinct variations to pick from.

If you want to expand a canvas, you can even leave the prompt empty and let the system guess the surroundings, though adding descriptors produces more accurate results.

In live demonstrations, the tool is impressive but clearly still a beta. While complex generated objects like cars or puddles can look slightly unnatural, it handles background textures and seamless blending remarkably well.

In some cases, it even carried over lighting sources from the original photo and placed accurate reflections on generated bodies of water, functioning as a much smarter, promptable version of Content-Aware Fill.

Commercial safety as the main selling point

The real breakthrough here might not be the pixels, but the legal clearance behind them.

Creative agencies have largely hesitated to use generative AI due to messy training data and copyright risks. Adobe claims Firefly is trained strictly on Adobe Stock images, openly licensed content, and public domain material without restrictions.

That means work created with Generative Fill is designed to be commercially safe out of the box, something rivals with black-box datasets cannot guarantee.

Adobe is also embedding Content Credentials, functioning like a digital nutrition label. This metadata travels with the image and lets anyone verify on verify.contentauthenticity.org whether AI was used in its creation.

Why it matters

This is generative AI finally meeting people where they actually work.

Stand-alone web tools were fun demonstrations, but production designers live in Photoshop. Integrating generation, inpainting, and outpainting directly into non-destructive layers eliminates hours of tedious cloning and masking.

We think Adobe's biggest advantage isn't raw rendering quality, but corporate trust. By offering commercially safe data and verifiable provenance, Adobe removes the exact legal barriers that kept high-paying enterprise clients on the sidelines.

This puts immense pressure on standalone AI image tools. When the reigning industry standard adds generation natively to the software everyone already pays for, third-party wrappers start looking unnecessary fast.

Sources (1)

  1. [1]
    theverge.com(opens in a new tab)Originaltheverge.com

Topic trackerAdobe's Generative AI Strategy

Every story on this topic, newest first.

  1. Adobe drops Elements 2025 with AI removal and depth blur
  2. Adobe Illustrator's vector AI actually works
  3. Adobe supercharges Illustrator and Photoshop with new AI tools
  4. Adobe brings Firefly AI and multi-doc insights to Acrobat
  5. Adobe brings Firefly-powered object removal to Lightroom

Best of AI, daily