While the world is preoccupied with ChatGPT, another AI revolution is taking place before our eyes. In only a couple of years, image-generating AIs have gone from laughable to mind-blowing, and they’re only getting started. Several tech companies have already unleashed powerful AIs and now, Adobe is the latest company to join the fray with their new Firefly suite — and just like several of its competitors, it’s pretty good.
It must’ve been a pretty frustrating year for Adobe. The company is used to being at the forefront of creative visual imagery, yet for much of the past year, it was forced to watch from the sidelines as the likes of DALL-E or Midjourney took the spotlight. But Adobe is finally launching its own image generator.
The company is releasing a “family” of AI models called Adobe Firefly. For now, they’ve published two tools from the suite — the first of them generates images from a prompt, similar to DALL-E or Midjourney, while the other generates stylized text, sort of like WordArt on steroids.
“We’re not afraid of change, and we’re embracing this change,” says Alexandru Costin, VP of generative AI and Sensei at Adobe. “We’re bringing these capabilities right into [our] products so [customers] don’t need to know if it’s generative or not.”
But Adobe couldn’t resist taking a jab at its competition. For starters, Adobe is one of the few companies that discusses what data its model is trained on. Furthermore, Adobe says everything in its model is copyrighted or licensed for training. So essentially, Adobe is saying it didn’t steal artists’ work to train its model, something that some of its competitors are accused of doing.
“We can generate high quality content and not random brands’ and others’ IP because our model has never seen that brand content or trademark,” Costin said. Costin also added that Adobe wants to pay artists to be able to train their model on their work. “We’re exploring multiple options,” Costin said.
Unsurprisingly, Adobe’s AI is also excellent at editing images — after all, this is basically the company’s bread and butter. While the tool is not yet integrated with other tools like Photoshop (or at least not publicly), having this powerful capability in an already popular software would clearly be very useful.
In a sample released by Adobe, the AI was shown to be able to generate stunning variations, such as changing the season in an image or generating different versions of a lighthouse.
Adobe also seems much more focused on ease of use than its competitors. Whereas start-ups were eager to demonstrate the working technology, Adobe wants to show how the technology can be used easily. For instance, its software features built-in options that can change the art styles and lighting. The product also offers a neat feature that enables the user to create a paintbrush based on something that’s in the image. All of this works both on generating images and on stylized text.
Adobe spent less time explaining how its AI addresses the other problems commonly associated with this type of technology, such as generating deepfakes or other types of unruly images though presumably (or hopefully), safeguards are in place preventing users from producing nasty forms of content.
The two tools are now available in a public beta. You don’t need to be registered with Adobe, but there is a waiting list.
The move also comes at an important moment for Adobe. Not only is generative AI booming, but the company spent a whopping $250 billion to acquire the design tool Figma. The move, which may face legal difficulties from the Department of Justice, was meant to cement Adobe’s spot as the king of creative tools, but you can’t really claim that crown unless you’re also strong in the AI game.
Adobe is now in the game. Whether it will come out on top remains to be seen.