Tuesday Adobe added A new tool for Photoshop beta called Generative Fill, which uses cloud image synthesis to fill selected areas of an image with new AI-generated content based on a text description. Adobe Firefly-powered Generative Fill works similarly to a technique called “painting” used in DALL-E and Stable Diffusion releases since last year.
Generative Fill is based on Adobe Firefly, Adobe’s own image synthesis model. As an AI model for deep learning, Firefly was trained on millions of images in the Adobe Standard Library to associate specific images with their textual descriptions. Now it’s part of Photoshop, people can type in what they want to see (like “a clown on a computer monitor”), and Firefly synthesizes multiple options for the user to choose from. Generative padding uses a well-known AI technique called “shading” to create a context-sensitive generation that can seamlessly blend synthesized images with an existing image.
To use a generative fill, users select the area of an existing image that they want to change. Once selected, a “Contextual Taskbar” appears that allows users to enter a description of what they want to see in the selected area. Photoshop sends this data to Adobe’s servers for processing and then returns the results to the application. Once generated, the user can choose from several generation options or create additional viewing options.
When using the Generative Fill tool, a new Generative Layer is created, allowing you to non-destructive changes image content, such as adding, expanding, or deleting, controlled by these text hints. It automatically adjusts to the perspective, lighting, and style of the selected image.
Generative Fill isn’t the only AI-powered feature added to the beta version of Photoshop. Firefly also allows Photoshop to completely remove parts of an image, erase objects from a scene, or enlarge an image by creating new content that surrounds an existing image.repainting.”
These features have been available in the OpenAI DALL-E 2 image generator and editor since last August (and around the same time in various homemade releases of Stable Diffusion), so Adobe is only now catching up with them by adding them to its flagship image. editor. Admittedly, this is a quick turnaround for a company that could have a huge liability target painted on its back when it comes to matters of creating malicious or socially stigmatized content, using artist images to educate data, and amplifying propaganda or misinformation.
Also, with Generative Fill’s ability to easily distort the apparent multimedia reality of a photo (admittedly Photoshop has been doing this since its inception), Adobe is redoubling its efforts. Content Authentication Initiativewhich uses content credentials add metadata to generated files that help trace their origin.
Generative seeding in the Photoshop beta app is currently available to all Creative Cloud members with a Photoshop subscription or trial in the Beta Apps section of the Creative Cloud app. It is not yet available for commercial use and is not available to anyone under the age of 18. not available in China and currently supports English-only text prompts. Adobe plans to make Generative Fill available to all Photoshop users by the end of the year so everyone can easily create backyard clowns.
If you are not a Creative Cloud subscriber, you can also try generative padding for free on the Adobe Firefly website by signing in to Adobe using the web tool. Adobe recently removed the waiting list from their Firefly beta.