In the midst of the AI revolution, a single question burned in my brain: Can free AI image generators create a photorealistic image based on the following prompt?
Cat picking daisies
Note that this devilishly difficult task includes three parts, two of which are not particularly difficult: cats and daisies. It is the action that proved difficult.
Nonetheless, the results of our testing were intriguing. Although image quality varied greatly, the range of imaginative responses is encouraging for the use of AI image generators as tools for creative expression. While all images included here are the first iteration, these can be considered a starting point, to which users can apply carefully crafted prompts to shape and edit the results.
Adobe Firefly
To use Adobe Firefly image generator, a free account is required. Provided are four aspect ratios, and dozens of styles, themes, and techniques. Users' free generative credit counts reset each month automatically, albeit with a somewhat complicated scheme for tracking those credits.
RESULTS: You don't want to meet any of these cats in real life.
BING IMAGE CREATOR
This AI image generator automatically detects if you’re signed into a Microsoft account, making it easy to start creating images right away. Start with 100 credits, then apply rewards points accumulated within your MSN account for more. Save, download, or customize your images. The platform applies controls to prevent potentially harmful image generation and includes a modified Bing icon in the bottom left corner of each image to indicate that the image was created using Image Creator.
RESULTS: Perhaps the most realistic set of images was generated by the Bing platform. Not only were the cat faces and flowers highly photo-realistic, but also the action depicted came close to the prompt challenge. To be fair, one must note that by depicting only closeups of cat heads, with no limbs or bodies, Bing narrowed the scope of difficulty. Perhaps this is part of its intelligence!
Craiyon
This relatively spare site is easy to use, but slower than average and offers only three style options for free users. To get faster results, buy a plan priced from $5-$20/monthly. Free results are automatically generated in a 3 x grid with nine images.
RESULTS: Craiyon’s algorithm had trouble with cat eyes, cat faces, and background specifics. However, give credit when credit is due: Ungainly as the results were, Craiyon at least attempted to portray the action of “cat picking daisies” with legs, mouths, and paws, instead of simply focusing on the nouns (cat, daisies) and avoiding the verb (picking).
Deep AI Art
Featuring an image aspect selector as well as dozens of free and paid styles to choose from, Deep AI also includes an AI chatbot. While the site is easy to use, images may provoke user responses from “Where is the cat?” to “Is that two bodies, one head?” to “Direct from the uncanny valley to your computer!” The so-called Genius Mode might do better, for $4.99 monthly.
RESULTS: One of four attempts yielded a photorealistic feline head, which may or may not have been attached to two bodies (you be the judge). On the other hand, photorealism is overrated. Users are advised to go with the slightly surrealistic flow.
Fotor AI Image Generator
To use Fotor AI Image Generator, create an account or login with Google. After the first free images, users will need to buy credits or a subscription. Choose from among five aspect ratios and dozens of styles. Need inspiration? Browse the impressive library of images generated by other users.
RESULTS: Three out of four images were fairly photorealistic representations of cats, grass, and daisies. Note that the software has the most trouble with eyes when the view is not head-on.
Imagine
The polished Imagine AI website boasts a six-million strong user base and 70 million images generated. The free image-generating options are limited to a 1:1 aspect ratio and eight styles.
RESULTS: High-quality flowers, cat heads, and faces, with somewhat disproportionately small bodies.
Night Cafe
Night Cafe offers five different text to image algorithms, with varying results. Overall, the highly imaginative approach is better suited to flights of fancy rather than photorealistic images. A robust community of users supports a chat room, strong image library, contests, and more. Users can accumulate a fairly generous credit package by taking actions such as publishing their images or liking other’s images.
RESULTS: From Cubist nightmare (disfigured faces and bizarre eyes) to women with cat ears, Night Cafe's AI-generated images are nonetheless not without charm.
Stable Diffusion Online
Fairly slow, but your place in the queue is always visible, as well as the expected time left. Fortunately, while the site proclaims “Freedom: No limitations on what you can enter,” graphic word prompts yielded a black image square only.
RESULTS: While the cats depicted are quite proportionate, detailed, and generally realistic, they are more akin to an illustration rather than a photograph. Nonetheless, if you have time to wait, the free version of Stable Diffusion is a good option to try. Pro Tip: Time of day is a big factor in how long you'll wait in the queue. Avoid rush hour.
