I Put Adobe's New Firefly Image Model 4 Head-to-Head With ChatGPT’s Generator—And the Results Are eerily Similar

If you've ever strolled through an art museum filled with various works, you may have observed that students from the same instructor or artistic movement often exhibit shared aesthetics and methods. Interestingly, this phenomenon also occurs with AI-generated images.

Adobe has been demonstrating its latest Firefly Image Model 4, which generates numerous striking visuals. Curious about how it compared with the recently improved image creator from ChatGPT, I put them both to the test. The outcomes led me to envision two budding yet highly skilled artists who have not fully established distinct individual styles. Instead, they closely follow their instructor's techniques, resulting in commendable but quite alike creations.

I tried out several prompts on both models to observe their outputs, and here are the results. On the left side, you'll find the creations generated by Firefly, while those produced by ChatGPT are displayed on the right.

Portraits

I began with a request for a portrait, as lifelike individuals are likely the most sought-after type of AI-generated imagery. I then instructed both AI tools to create one. A detailed shot of a redhead with blue-gray eyes and subtle freckling, adorned with a straw hat, positioned within a sunflower field during the golden hour.

These two digital avatars appear as though they might be close relatives, possibly even brother and sister. Their lip shapes, hair textures, and even the contours of their hats are almost indistinguishable from one another. While Firefly’s “photograph” seems to depict a scene set on a hotter day due to the clothing choices, the similarities between them remain striking. Identifying which artificial intelligence model generated what image would prove challenging without additional information or data, yet both models evidently underwent training using photographic archives containing images of some overlapping individuals.

Complex Scenes

Next, I was curious about their performance in a intricate, chaotic scene. Since I am an admirer of B-movies from previous eras, I requested both AI tools to create A science fiction lab in disarray, featuring shattered glass, intermittent lighting, frantic scientists, faulty robots, and luminescent extraterrestrial samples breaking free from their confines.

There were significant differences observed here. Intentionally, several distinct events were woven together simultaneously. "Firefly" centered around the robots, whereas "ChatGPT" highlighted the alien element, even though some of these aliens might also be robots. Each succeeded admirably in populating the scene with various activities from differing perspectives. However, I would lean towards "ChatGPT," primarily due to the inclusion of alien minds and how the scientists are frantically fleeing for safety.

Breakfast time

Feeling a little peckish, I next decided to see how the two would perform at making me a (digital) breakfast. Food can be tricky to get right with AI, and it's always very obvious when errors creep in. That's why I wanted to see how they did at showing me: A bird’s-eye look at an exquisite morning feast featuring pancakes, berries, and a latte adorned with latte art.

Both models were quite enthusiastic about showcasing their berries. The alternative Firefly choices resembled a truckload of produce overturned onto the dish. Meanwhile, ChatGPT seemed to assume I’m akin to an average young kid who would gleefully eat berries till bursting. Despite this, the pancakes appeared appetizing in each case. However, the issue resurfaced with the latte art. Although I hadn’t specified a design, it was clear that fern patterns are common; interestingly, both featured a heart atop the fern shape, although one of Firefly’s cups evidently had some cinnamon spillage beforehand.

Legendary words

I chose to add some extra details to enhance the prompt further. The concluding version requested both Firefly and ChatGPT to create. A magnificent dragon adorned with shimmering scales was seated on a snow-capped mountaintop during sunrise, its enormous wings completely outstretched. It exhaled a jet of flame that coiled into the phrase 'Dream Big' in vivid, blazing letters.

With countless depictions and descriptions of dragons available, all I specified were shimmering scales and an awe-inspiring presence. Surprisingly, both Firefly and ChatGPT generated dragons featuring identical horned frills around their faces and tail structures reminiscent of dinosaurs. Using the analogy of an art student, one might say that Firefly paid more attention to the scenery rather than focusing on intricate lettering.

ChatGPT, on the other hand, interpreted the bold and vibrant lettering in the prompt quite literally, resulting in something akin to clip art for Firefly’s design. However, the dragon's posture appears much more dynamic and engaging. Nevertheless, both could just as easily be two aspiring artists receiving guidance from an instructor about how sunlight manifests during sunrise amidst mountainous terrain.

Firefly Image 4 and ChatGPT both have the capability to produce stunning, bizarre, or eerie visuals. However, distinguishing between these two systems might be quite challenging. They are skilled at crafting human visages; indeed, their creations sometimes appear as though from the same lineage. Additionally, they excel at depicting bustling scenes, spelling out words using flames, and even preparing meals.

These models and similar ones have now advanced to the stage where choosing one often depends more on your budget and compatibility with other items from the same brand than on factors such as picture quality or personal preference.

You might also like

• I pitted ChatGPT's updated image generator against DALL-E 3, and it represents remarkable progress—though it does require some endurance.
• I brought my doodles to life using Deep Image, and it feels just like wielding a magical paintbrush.
• I'm not going to use ChatGPT’s Studio Ghibli image generator because it contradicts what I appreciate most about those films.

If you enjoyed this article, click the +Follow button at the top of the page to stay updated with similar stories from MSN.