I decided to try out the Beta for Adobe Firefly, which is Adobe’s new AI art tool. The Beta is free for anyone to try right now. None of the images can be used for commercial purposes while in this Beta stage. I think they are testing things to see what people would use it for.
I tested it out with some squirrel related pictures for the purposes of this blog. I wanted to see if it would draw a squirrel drawing or painting something.
Initially the results were rather strange, however I tried different colour squirrels and discovered I could define the background colour as well. At times the squirrels tails or limbs seemed to fall off. Perhaps squirrels were too much for it I thought.
I persisted with the experiment and tried different word combinations. It seemed that as long as I didn’t put too much into the scene it worked better, but I got some crazy results as well. I have no idea why there is a cats hand drawing in the first picture here because I didn’t ask for that. In one of the pictures a squirrel drew another squirrel instead of a nut. Sometimes they would start eating the nut or try to paint onto it rather than paint a picture of it.
I tried to get the squirrels to wear trousers and different types of hats as well. It didn’t really understand what I wanted with the beret but it did beanies and top hats well. I would say the trousers were a bit of a disaster but it would be hard to get a real squirrel to wear clothes so I persisted with this instead.
Eventually I managed to get the squirrels to do what I wanted by changing the words. Each result is completely random and sometimes it worked and sometimes it didn’t. I tried many more pictures than the ones I’m showing here, these are just a sample of the better results.
Here are some of the best results I got. None are perfect of course, hands are off, tails are a bit all over the place. I think it is cool how you can change the colours and things the squirrels are doing so fast though. It is great for inspiration. Obviously I would rather paint it myself so I get it exactly the way I want it if it was going to be for a final image, but this could be a good starting point for reference if I was going to do so.
As well as squirrels I also tried the lettering tool. This tool fills in fonts with images of what you type. In my opinion it may be useful for inspiration but if you look closely at the details they are not great. It performed better with inanimate objects rather than animals.
I also tried making some mermaids and elves. Nice colours and it recognizes different ethnicity’s.
As a bonus here are some more images I tried.
I couldn’t get it to do female squirrels unfortunately.
I figured out the only way (as far as I know) to get girl squirrels is to put them in a dress. I think that is a little sexist bias on the part of Adobe but maybe they will fix it. Of course I am dubious about the censorship and bias that could be with these tools, but they are probably aimed at a primarily commercial audience being from Adobe. I found this process a bit tedious this time around because it took a long time to get nice results as well. It isn’t much fun pressing a button, I’d rather draw in all honesty, but again this is a nice tool to try different colour combinations. For anything specific or complicated it falls apart at the seams.
Sometimes the squirrels in dresses looked rather odd.
Anything other than a dress or skirt turns the squirrels into a boy (albeit a pretty one).
I hope that was interesting.
Thanks for reading.