If you have been paying attention to the recent NFT boom and bust, you may have come across a collection called Bored Ape Yacht Club. This is a collection of profile pictures of cartoon apes generated by an algorithm. Only minor changes in colors, clothing and expression are made to each ape to form a new unique NFT creation. Tens of thousands of similar art works were being created using the algorithm in just seconds. This inspired others on the Ethereum blockchain to do the same and soon the platform was flooded by these algorithm generated images leaving actual digital artists who spend weeks on a single art piece in the dark.
A Non-fungible token was to help artists receive royalties for the work they have done even after multiple sales of the art work. In a way the Openseas platform and the NFT concept was supposed to inspire and encourage artists to make a living off of what they do but were overshadowed by few bad actors. Artists were outraged and there were many debates on whether algorithm generated art is real art? The question still remains relevant.
Recently there’s been another innovation that’s gaining some attention among digital artists. Its called text-to-image generation using machine learning. Since mid 2010 s, there was some good research and development being done in the field and in the past few months it has got to a point where the AI has started spitting out photo realistic images in just seconds. This time however, rather than being threatened, I was actually amazed by it’s capabilities.
Let’s have a look at some of the text-prompts and what image the AI generated and judge how well they have done.
I have used a free AI image generator from Picsart. The Images are as generated and no additional modifications were made to them.
Picsart provides a number of artistic layer options that can be added on to get the output closer to what you are looking for. For most images I have selected Highly detailed, Ultra-realistic, 8K, Hyper-realism, Realistic lighting, Photo-realism, Ultra detailed, Fantasy and Unreal Engine with just slight variations for may be a few.
Prompt 1: Mermaid in green water
Prompt 2: Girl eating ice cream
Prompt 3: Bird flying over city
Prompt 4: Dancing girls
Prompt 5: Mermaid in water
Prompt 6: Princess with dragon
Then, I decided to take it a little further to check how imaginative the AI really is.
Trying out a few complicated prompts:
Prompt 7: Vin Diesel and Ryan Reynolds
Prompt 8: Gal Gadot as Superman
Prompt 9: Gal Gadot as Spider-man
Prompt 10: Tom Hardy James Bond
Prompt 11: Margot Robbie as Jack Sparrow
Prompt 12: Elon Musk living on mars
Prompt 13: Apple Car
Prompt 14: Tom Cruise relaxing
Alright, now to conclude, let’s go back to a simple prompt.
Prompt 15: Girl at museum
I’m impressed by how the AI is aware of the fact that at a museum there would be works displayed on the walls and decided to scribble something onto the wall to give it a museum feel. This might just be my favorite of all the images it generated.
Machine learning used to generated these images relies on the images it scrapes from the web. A lot of times you may feel like you have seen these images before.
As this technology further advances, it may affect the livelihood of traditional digital artists and photographers. There have been calls for AI artists to declare their works as generated by AI when on sale but it remains to be seen how far that goes.
Use cases for this technology could be quiet diverse but would be very helpful in pre-production works for movies and TV-shows, Illustrations for newspapers and publications and of course art works.
As for the software engineers, coders and mathematicians that have spent many years developing these neural networks for machine learning, I appreciate and acknowledge the work you have done.
You can try Picsart Image generator for free using this link.
If you’ve enjoyed this article, be sure to share it with your friends and family.
