Artificial intelligence has reached the masses.
You can say that ChatGPT has become old hat by now. Everyone knows it. In the meantime, there are even discussions in Germany about regulating the program. The pictures of Pope Francis in streetwear clothes or Donald Trump in handcuffs have gone viral and raise many questions.
That's because both images were created in just a few minutes using the program "MidJourney." They look so real that at first glance they can hardly be distinguished from reality. Midjourney is a paid "text-to-image" AI (with free trial) that can produce detailed images, of any style and format, based on written instructions.
There are also free programs, such as the Bing Image Creator, Dall-E, or "Staple Diffusion". Except for the Bing variant, however, this requires either a certain basic IT knowledge, an elaborate computer, or a lot of patience, since the images can hardly compete with Midjourney in terms of quality.
At the moment, however, it is proving difficult for all of them to create print-ready illustrations, to edit the image outputs and to combine different works. In addition, the question of image rights is still vague: To what extent artists are compensated whose works an AI draws on as a reference remains unclear.
Now Adobe, the market leader in creative programs, is trying to fill exactly this gap and is launching "Adobe Firefly", a software that seems to be worlds ahead of all other AIs.
What it's all about - robotspaceship clears it up!
What is Adobe Firefly?
Adobe Firefly is a multifunctional AI tool. It is currently still in the beta testing phase and can only be used via an invitation link.
Like other text-to-image programs, it accesses a database of reference images that is invisible to the user but immeasurably large, and works its way through the database to implement an instruction as accurately as possible.
So if I want to see Pope Francis in street fashion, the AI goes through my instructions piece by piece and compares a huge number of images of the Pope, streetwear clothes, locations in the real world and pulls small fragments from all of them. Since these fragments can always be different, it is almost impossible to get exactly the same image generated twice, even if the instruction remains identical.
Now, based on the fragments, the AI assembles a logical composition of the image, usually in a few seconds, and plays it out to the user. So far so familiar.
What makes Adobe Firefly so special now?
A great image is all well and good, but most of the time you create graphics to make money with them. You could sell the graphics digitally as JPG, PNG or PDF, but you can't print them on objects like shirts, posters or banners.
Usually, for such purposes, so-called vector graphics are needed, which keep the same resolution and quality, no matter how they are drawn and stretched. So far no AI can build these files - except now Adobe Firefly.
You can chase the results of other programs through "image-to-vector" software. But the results are nowhere near as good or intuitive. Firefly is already a pioneer in this respect.
Post-processing of already existing images
Images often don't just want to be created. Post-processing is one of the biggest hurdles before publishing an image. Many programs, such as Adobe's own Photoshop, are more than adequate for retouching - but they also require some basic knowledge.
If you want to change an entire outfit on a photo, laymen usually get into trouble. In the teaser of Firefly, however, you can see how a red leather jacket is put on a model in less than a minute. The result is completely without any editing traces - no blurred backgrounds, no prominent pixels or altered proportions.
What's particularly exciting here is that you can select the area you want to change so that only this area is taken into account by the program.
Alternatively, users can also convert the AI-generated images into different styles at the touch of a button. In all competing products, one would not only have to know the names of these styles beforehand, but would also have to enter the entire instruction again. In this case you risk that the images are changed completely and not only their style.
Creating your own fonts
Especially for creative design fonts are very important. Often you have good ideas, but it is difficult and tedious to convert them into a typeable font.
Now Firefly can create the most fancy designs based on a description, which can be ported and used in other programs in no time.
Merge multiple images
If the birds are missing in a beautiful landscape photo, or the one person who missed the appointment is missing in a family photo, thanks to Firefly you now have the option to realistically merge several pictures into one in a flash.
Of course, you should also own the rights for the works you merge here if you want to use them commercially. Adobe mainly does not use its own databases here, but your input.
Image rights and compensation
The question of who now owns the AI-created images is always a topic of public debate. Due to the fact that AI is still very young, no clear decisions have been made about this at the moment. Roughly and without guarantee, one can say that the images of most AI's can be used commercially. However, the rights themselves lie with the creator, i.e. the AI.
When creating content, however, the program also accesses a database and gathers inspiration from the works of other artists. Depending on how much an AI draws inspiration, legal problems could arise here. And there's another factor: currently, these artists of the original images are not compensated - even though their images and their styles are sometimes used thousands of times on the Internet.
Adobe Firefly solves all these problems by simply using its own internal database. This is much smaller than that of most of its competitors, but at least it provides legal security. Also, the artists whose work is used can be tracked and paid for it.
The smaller image database, which Firefly uses, does have a catch: For some inputs, there are simply too few reference images to guarantee a similarly good output quality as with Midjourney. Sometimes faces are slightly washed out and house facades appear crooked. However, most outputs are able to keep up with those of the competition.
We live in a world where one disruption follows another. Digitization, AI, metaverse. Who is supposed to keep track of it all? The answer is us, the robotspaceship crew. We help you stay on top of things and make your business fit for the future - whatever it may look like. Your resilience is our mission.
Adobe Firefly has a clear chance to become the text-to-image market leader for businesses. It has plenty of exclusive features, seamless integration with Adobe's Creative Cloud, which is already used by many companies, and is perfectly legally protected for the future.
It only remains to be seen how Germany, or even the EU, will evaluate these technologies, their privacy and legal issues in the future. It is likely that a lot will happen here, especially in copyright. This could cost many AIs their clear database advantage over Adobe.
Jason Schmitt is a german digital media specialist, experienced in VR/AR applications, social media and content creation. He is an award-winner of international competitions in the Design, Marketing and Sales sector.