I decided to give Adobe’s Firefly image creator a try. It may not be flawless, but it’s a wellspring of creative output Additionally. There are new features, such as generative AI in Illustrator and Adobe Express.
Adobe has made significant advancements to its Firefly family of generative AI tools, aiming to boost creative productivity. These updates include a substantial upgrade to the engine behind Photoshop’s image generation and the introduction of generative AI for Adobe Illustrator designs and Adobe Express layouts.
Firefly, known for turning text prompts into images, is already accessible on the Firefly website and within Adobe’s Photoshop image-editing software. The latest AI model, a second-generation version, promises improved image quality and more intricate details. While it’s not available in Photoshop just yet, Adobe plans to roll it out after some online testing.
In my own testing, I observed considerable improvements in both results and user interface. Adobe trained the new AI model with double the number of images, resulting in higher resolution, better detailing (down to skin pores), and the ability to control image generation using photography parameters such as lens focal length and depth of field.
Generative AI relies on established AI methods that recognize patterns in real-world data. However, it takes things a step further by generating new content, like text, images, speech, or video, based on its comprehension of these patterns. This advancement has transformed the capabilities of computers. Iiberating them from rigid, literal processes and offering a glimpse into the potential of truly intelligent machines.
Testing Adobe’s enhanced Firefly generative AI was a delight. Although the results may occasionally lack convincing realism. The process of creating imaginative and entertaining images is immensely enjoyable, particularly when working with forgiving art styles like paintings, cartoons, and watercolors. Typing in prompts, observing the AI’s creations. Tweaking the prompts for more practical or outlandish results was a creative adventure.
When testing AI, one of my go-to prompts is “a parachuting elephant.” I observed that the results from the upgraded Firefly were notably better than its first-gen counterpart. Although the technology still grappled with certain details, particularly the parachute cords. In one instance, it even managed to construct a wooden frame to support the elephant, which was quite surprising.
For the prompt “a kindly doctor in a hospital room for a pharmaceutical ad,” Firefly delivered various generally acceptable portraits. These portraits depicted diverse races and genders. But they all featured the standard medical scrubs and stethoscope, meeting the expected visual criteria.
As for the task of generating images of a “spikey electric guitar,” Firefly excelled in producing visually striking and edgy representations. Even if they occasionally lacked the precise number of frets, pickups, or strings. wellspring of creative output
However, my testing also underscored that there is still significant room for improvement. In images of the stereotypical hacker hunched over a keyboard, some had their hoodies worn backward. Indicating occasional quirks in AI-generated output. Unfortunately, my quest for a photorealistic red crab waving its claws in the air remained unfulfilled, with distorted claws, extra eyes, and other issues. Firefly managed a passable representation of a groundhog. But it appears that Adobe’s training data might not encompass the more uncommon pica, a rodent found in high mountain regions.
The interface has also seen improvements. Users can now select between photorealism and a more artistic illustration style, with options for various aspect ratios like square, landscape, and portrait. Sliders for visual intensity and style strength provide choices ranging from dramatic to understated. And users can even upload a reference photo to influence the style of the output.