And the outcomes are quite favorable. Adobe’s new AI image generator, Firefly, is making waves.
Similar to Midjourney, but with various variations, both in terms of how it functions and also the direction that Adobe is moving this AI platform in its plans. Let’s have a little discussion about it.
Regarding how it functions, you have already seen that it is possible to enter text and produce images. The right half of the room is dedicated to variables that can be applied to the creation of images. For instance, we can alter the aspect ratio. Perhaps you want a widescreen. Firefly will now produce photos with that aspect ratio, sixteen by nine.
Additionally, you have a choice of Content Types and Styles. Here are a few well-known ones, as well as lighting, composition, and Color and Tone. We could, for instance, switch to dramatic lighting, modify the composition to have a shallow depth of field, and select Generate to examine the outcomes with the updated settings.
Go to firefly.adobe.com to access Firefly. There is a waiting list because it is currently under beta testing. Once inside, you can see the two options available to you. What you just saw was text to image, also known as text effects. You may create text effects that look like these by clicking Generate.
How does that function? Very simple. You simply type your text. I’ll type my initials in this instance: J and R. I’ll explain what I want my Text Effect to produce along with myself. It might resemble Golden Ornaments in some way. That text will be produced in Firefly using the description and the reference text, effect.
Sample Effects are seen on the right side. To see them all, you can expand the images. You may make cookies, flowers, or color marvel. Utilizing Firefly has the benefit of allowing you to have access to options like Text Effect Fit. How well does the effect match the words, then? Will it fit snugly or loosely?
The fonts in this list are another option. You can view everything by selecting “View All.” You can click on the typeface you want to use from among all the options. I believe a tight fit will be more advantageous in this situation. You can try with any prompt you like, of course. Consider Grass and observe what sprouts.
Then, you can pick from the several Styles. Four styles are displayed below, and you can click on them to add that style to your text. Additionally, perhaps you should try out the capability to pick a background transparency, which is another significant aspect. After you’re prepared to export, you can click on this background transparency. By pressing the Download button, a PNG with a transparent backdrop will be exported. Alternatively, you can use any hue if you’d like, export the picture as a JPEG with the background color that you like.
When this recording was made, these two choices are available to you with Firefly, which is now in beta. But as you can see, Adobe intends to produce more AI utilities, such as a vector recoloring feature. These instruments are being used for exploration. The Text to Brush feature, where you can create brushes, is possibly the most intriguing to me. Brushes for Photoshop using a full description.
Currently, one of your inquiries might be, “How is it different from Midjourney?” Of course, one of the greatest distinctions is that Adobe will integrate Firefly into several of their artistic applications, like Photoshop. Another striking distinction is in how Adobe promises to carry out the Adobe Firefly system’s development of image-generation techniques.
Adobe claims that Firefly will prioritize business use. Adobe’s plan involves instructing Firefly using Adobe Stock, along with other materials that have been licensed or have expired copyright. To be crystal clear, this is content from Adobe Stock, not from your Creative Cloud library, your own work, or something similar.
According to Adobe, you’ll be able to instruct Firefly using your voice in the future. Additionally, Adobe claims that it is investigating contributor pay models that will enable them to make money from the Firefly training content. Although there aren’t any specifics on how that will operate, it’s good to know that Adobe is considering distributing earnings to the users who are teaching Firefly.
What’s better at this point, you’re probably asking yourself? Midjourney or Firefly? There is no doubt in my mind. At this point in the adventure, it is considerably better. It is trained with a lot more material, a portion of the reason being that Adobe has copyrighted the content being trained on, employing only materials for which they have the copyright.
Midjourney has undoubtedly existed for a little longer, but just as long as necessary, I’ve used the same prompt as an example on both Firefly and Midjourney 5.
And the outcomes are as follows. Midjourney is on the left, while Firefly is on the right. You may see the prompt I use below. Generally speaking, I make an effort to keep the prompt simple, with no additional settings or tools, simply to test what an easy prompt from the A.I. might produce.
Of sure, you can achieve greater outcomes simply utilizing such add-ons on either of these applications.
In this initial comparison, I chose a studio image featuring an adorable creature with large green eyes. Both Midjourney and Firefly demonstrated impressive performance in this scenario. Personally, I lean towards Midjourney as the winner due to its refined depth of field and other visual aspects. However, Firefly also exhibited commendable results.
Next, I shifted my focus to capturing the essence of a golden hour with tacos and avocado on a wooden table, backed by the New York City skyline. Firefly’s potential superiority in food photography stems from its training with food stock images on Adobe Stock. Nevertheless, I found that Midjourney surpassed Firefly’s performance by delivering a more appealing background and foreground composition.
It’s worth noting that Firefly encountered challenges with certain backgrounds, as evident in the top left image.
Exploring different scenarios, I attempted a realistic 1980s Polaroid photo featuring a man standing in front of a muscle car. Firefly encountered challenges once again in this context.
It’s as if Firefly hasn’t had its fill of experience yet. In contrast, Midjourney hadn’t encountered Polaroids before, but it managed to generate remarkably authentic-looking Polaroid images.
In this instance, I’d give the edge to Midjourney. It delivered what I sought with finesse. However, there’s a peculiar aspect to the man’s scale and his suit, which adds an intriguing touch.
Next up, a close-up of an ant’s head, and both Midjourney and Firefly proved their prowess. Midjourney’s rendition of the ant’s head was spot-on, capturing the essence of the subject. On the other hand, Firefly’s contribution included captivating macro shots, showcasing its versatility.
Considering a query from Firefly about product photography, I turned to Adobe Stock’s collection of objects on plain backgrounds. The “blue headphones” prompt yielded impressive results, with Firefly excelling against a solid yellow backdrop.
While Midjourney performed admirably, Firefly’s specialization shone through, offering more aligned outcomes with the prompt. Its proficiency in this context was undeniable.
As often happens, these AI-generated results align with their training data. Firefly’s adeptness in certain domains is evident, although Midjourney’s overall performance remains impressive.
The experimentation continued with a twist, as I used a nighttime shot of a muscular cat lifting weights in a tracksuit on a rooftop against San Francisco’s skyline. Neither AI performed exceptionally well in this instance.
Regarding the previous comparison, Midjourney’s overall performance seemed better to me. However, the image on the right appears to be more in line with my original prompt’s intent.
If I were to choose one of the eight images, I’d go with the firefly picture, despite my belief that Midjourney performed better overall.
Continuing the experimentation, I embarked on another whimsical endeavor: a steampunk character sporting red goggles, riding a unicycle, and waving. In this instance, Firefly’s output slightly outshone Midjourney’s.
It’s logical that the results resemble stock photos, given the AI’s training data. While I received a few “waves” from Firefly, Midjourney provided just one wave without the red shorts.
Although red is the dominant color, the shorts were missing. Speaking of the unicycle, I only got a single image from Midjourney. Perhaps two were generated, but I’m confident about one.
Next, I turned to a pack of adorable dogs, captured in a selfie during sunset’s golden hour at a park. Undoubtedly, Midjourney emerged as the winner in this scenario.
However, there were issues with Firefly’s results, including distortions that weren’t quite pleasing. The majority of the images produced by Midjourney showcased decent quality.
Then came an attempt at a flat-design logo for a towing company, featuring gray and orange as the main colors and a tow truck. Both AIs excelled in executing the prompt’s color requirements.
In the end, they both offer designs that I enjoy. Personally, I’m slightly more inclined towards the Midjourney design.
It’s noteworthy that Midjourney had a notorious struggle with producing hands, but Midjourney 5 managed to improve that aspect.
In this particular scenario, Midjourney demonstrated impressive performance.
Kudos for the well-timed utilization of a lone human hand against a pure white background.
From my observation, there’s only one hand visible in the image.
Interestingly, the finger located at the top right seems to have more than five fingers.
Regarding the four images generated by Firefly, we achieved quite a satisfying outcome.
However, Midjourney presented me with a white background in contrast to Firefly’s gray background in all images.
Shifting focus, I concluded my experimentation by questioning Firefly and Midjourney to produce AI-generated art.
In my opinion, both of them managed to deliver commendable results.
While these AIs might not consistently produce works of art, the intriguing nature of the photos cannot be denied.
Personally, I find myself leaning towards the greater variety provided by Firefly.
Even though Midjourney’s photographs are of high quality, they tend to share a similar aesthetic.
As the topics were identical, I’ll concede this round to Firefly.
However, as I mentioned earlier, using modifiers and other options can significantly enhance the results.
Either of these applications has the potential to yield even more impressive outcomes.
Feel free to share your thoughts in the comments section below about the findings of this comparison.
Artificial Intelligence undoubtedly plays a substantial role.
While debatable, Adobe’s involvement was expected and is worth discussing.
Share your opinions in the comments section below.
Let us know your perspective on Adobe Firefly and how you feel about its capabilities.
]]>