The software house Adobe is planning on adding Firefly, its newly introduced generative AI, into its flagship software later this year.
Adobe’s recent launch of its own AI, Firefly, has been quite successful. While the images it can produce are quite impressive, it’s the further plans the company has for it that are most intriguing. The first of these plans, adding Firefly into Photoshop, is happening much faster than people expected.
A future version of Photoshop to be introduced later this year, will include “Generative Fill”. In demos given to the press, Adobe showed that it could introduce new features to images that never existed before.
For example, a snowy landscape in a portrait position can be expanded on to create an entirely new image from its database. A dog running down a lane can have bubbles and a puddle added, or a car can be turned into a cloud.
Adobe already offers a few cloud-based features that integrate with its Sensei software to alter backgrounds, change faces and also use its “Content Aware” fill.
This functions differently from the “Content Aware” already on offer, as it is literally generating something that doesn’t exist based on the information given, rather than masking or adding.
Subscribe to our newsletter for the latest updates on Esports, Gaming and more.
We tried Adobe’s Generative AI
Generative Fill is now included in the web version of Firefly, and we gave it a go with an image from an upcoming review. We wanted to see how far we could push it and found that, despite impressive tech, it’s still clearly a work in progress.
It did, however, allow us to change the shirt, background, and beard.
Adobe adds new info to metadata to combat fakes
To get around the forgeries and ongoing concerns with AI-generated images after a few went viral, Adobe will be lading the image metadata with information about Firefly. This should give those online a good indication to decipher fakes. Adobe also currently watermarks images made with Firefly.
Adobe also plans to remove the waitlist for the online tool at some time in the future, while also keeping it free. Since March, the company claims that it has seen over 100 million images generated via the service.
Unlike Stable Diffusion, which can be run locally on your own system, Adobe will be operating similarly to Bing AI and DALL-E 2’s online generators. That means no celebrities or excessive images can be generated from prompts, and will simply give you a message that it failed to do so.