Adobe’s Artificial Intelligence Model, Firefly, Coming with a Rumble

Firefly, an artificial intelligence model produced by Adobe as a competitor to Midjourney and DALL-E, was presented to users as part of beta testing. Although Firefly is successful in some areas, it still needs improvement in many areas.

With the rapid rise of artificial intelligence models in recent months, companies have started to make investments in this field and offer their own artificial intelligence models to the market one by one. Among them, the most striking company of recent days is Adobeit happened.

Firefly, Adobe’s new artificial intelligence model in testing, is still under development. ambitious enough to challenge many visual artificial intelligence models . However, it can also fail in many areas.

It fails to model people while producing much higher quality and sharper visuals.

Image created with Midjourney 5 (left) and Adobe Firefly (right) using the same prompt

Unlike other artificial intelligence models of Adobe such as neural filter and smart object selection, Firefly allows you to create images from scratch instead of manipulating them. Working in a similar way to its competitors such as DALL-E and Midjourney in terms of usage, Firefly can offer users much higher quality and sharper visuals. Although artificial intelligence, which is still in the early testing phase, creates very detailed and successful scenes in some images, in some cases can create absurd or too simple visuals.

The majority of users who tested Firefly stated that the system was successful in creating animals and environments; it turns out to fail almost every time the system is asked to paint people. The main reason for this is that this artificial intelligence tool takes advantage of Adobe’s own stock image pool. While there are mainly landscape and animal images in the pool, It is stated that the images with people are very few.

Firefly also focuses on protecting the rights of working people

One of the most criticized aspects of DALL-E and Midjourney is that these artificial intelligence models develop themselves by scanning visual sites, social media and portfolios on the internet, especially Shutterstock. Naturally, this means that the images of the people who share the art they produce on the internet. It causes unauthorized use and similar images can be created using artificial intelligence systems..

Adobe, on the other hand, explains that it has taken multiple measures to avoid a similar problem with Firefly. For example; As we just mentioned, Firefly trains itself with royalty-free images only in Adobe’s image repository. In this way, the content of content producers who share their work on the Internet is not used without permission.

In addition, in the Metadata and codes of all images created with Firefly It is clearly stated that the image was created by the artificial intelligence system . In this way, people can easily distinguish between images produced by artificial intelligence and those drawn by hand.

Adobe also states that design companies can customize and develop Firefly using their own portfolios. As the improvements made will be specific to companies only, copyrighted images of companies can also be protected.

It is not known when Firefly, which is still in beta, will be available to all users, but if you want to sign up for the beta located hereYou can use the link.