Last modified: September 26th, 2022
AI is super cool, and also kinda scary.
It can create hilarious predictive text works like Harry Potter and the Portrait of What Looked Like a Large Pile of Ash or even surprisingly convincing Taylor Swift songs. It can also create disturbingly realistic deepfaked images or videos of famous people.
But also, AI can make art! I’ve only recently discovered some AI-based tools that can be used to create impressive creative works, which can be inspiration for artists and a supplement for skill for non-artists, allowing everyone to translate ideas into images.
These are my favorite, FREE AI art creation tools that I use to inspire my own works.
NightCafe Creator is a service that allows you to generate images using text prompts. It can also apply the style of one image onto another, but I find that feature less exciting than the generation of a brand new image from nothing.
NightCafe gives all users 5 free credits per day. Each image costs at least 1 credit to generate, and costs more if it’s higher resolution, larger, etc. You can also earn free credits for earning achievements, such as earning a certain number of likes on your works.
Generated images are usually pretty abstract, with vaguely recognizable components of familiar shapes smashed together in bizarre ways. I would personally never come up with colors, shapes, or ideas like this on my own. It’s almost like touching that childlike unrestricted creativity that so many artists strive to get back to in adulthood, but even more surreal and hallucinogenic. Sometimes there is a clear figure-ground relationship, a Euclidean space that you can mentally parse, but often it’s just a dismembered mishmash of almosts. That’s almost a cat. Almost a human hand.
I have yet to experience an image that doesn’t “feel” right, that doesn’t somehow capture the essence of the prompt.
The generated images are better when they defy your expectations – sure, you might go into a text prompt with an idea in your head, what you would draw to match that prompt, and then the AI does something completely bizarre with it. It breaks down well-trodden mental paths and lets you see a new take on an idea.
It’s almost like being able to commission the human collective unconscious for artwork.
But it’s kinda weird how obsessed with neon pink it seems to be.
NightCafe now offers Stable Diffusion, the gold standard AI art model that makes amazingly coherent shapes.
Dream by WOMBO
Dream by WOMBO is a simple, free, no-login-required AI art generator good for quick idea inspiration. The results are usually even more abstract than those of the above NightCafe, but it’s great to get a quick idea for composition or color palette from a text prompt. All you have to do is enter a text prompt and select an art style and you’ll have an AI-generated image in seconds.
Oh man am I excited to put this option on the list. DALL·E is one of the best “quick” AI image generators, creating NINE different iterations of a single text prompt at once. These images are all pretty low resolution which makes it quick to generate, and it is amazing for prompting ideas. It is simpler than some of the other tools on here – you cannot select art styles or ask it to work from an existing image, but it is the best at quickly spitting out lots of cool variations of an idea.
This tool is great for generating coherent objects, but not for more painterly abstract compositions. Most of the images generated here will be clear figure-ground objects on a white background.
I think these images really speak for themselves to express GauGAN 2’s power. This simple MS Paint-like interface can be used to roughly doodle features like mountains and trees, and then with the press of a button, it is turned into a photorealistic landscape.
It is astoundingly smart at interpreting vague blobby shapes scribbled in with my mouse – it understood that my vertical line of “water” against a background of “mountain” meant that I wanted a waterfall.
You can also switch the climate, time of day, and general feel of the generated piece by using the style filters in the top-right, and upload your own.
It doesn’t accommodate non-Earthlike landscapes; you won’t be able to make floating sky-islands or mushroom forests. Sometimes things like trees look rather wonky, and it’s best to simply scribble those in and not worry about drawing a realistic tree.
It’s also just FUN to use and play with, and the AI-generated results are nearly instant
ArtBreeder lets you literally do “selective breeding” with generated images, allowing you to create images based on other generated images to get closer and closer to your ideal. You can start off with nothing, saving a few images that vaguely capture some aspect of your desired output, and then “breed” those images together to get an even more desirable output.
Each press of the refresh button gets you 3 images based on your input – whether that’s the default “genes” (extra settings that affect output, like “long hair”) or whether that means you have the maximum of 8 possible parent images. You can use any of your previously-saved “children” or snag popular ones from the community. You can just hammer on refresh until it generates an image you like, or, if it keeps failing to create something you like, you can tweak the parents or genes.
I particularly love the Portrait generation category for envisioning faces of original characters. Also, I am so pleasantly surprised that you can generate furries. I just wish every category had an “upload” feature to freely combine any two images.
Text to Pokémon
This model requires a free GitHub account to use.
A variation of Stable Diffusion trained exclusively on Pokémon, it creates astoundingly believable ‘mon designs from a text prompt. It’s not the fastest image generator, but the results are worth the wait.
Here’s a result for “wendigo cryptid”, trying to generate creepy/horror themed Pokémon!
I hope you make use of some of these favorite tools of mine to create something awesome! It’s very fun and even a bit addictive to play with AI generation, and I find it tremendously inspiring for my own artwork. Let me know if you create something you want to share, or find another AI tool you like!
As an addendum – I do NOT support AI art models as a replacement for artists. I do not support AI-generated art being accredited to the person who typed in the prompt. I do not support AI-generated art being submitted into art contests or galleries. Typing a prompt is NOT the same as drawing an image. AI-generated art is NOT the creative work of the person who typed in the prompt and should not be regarded as such. AI art should also never be used to try and replace or mimic the work of real artists, or to try and mass-produce art in a particular artist’s style. Support real artists! Use AI for fun and inspiration and launching points ONLY!