Roughly five months ago, I held an AI art competition related to the fantasy world I’ve been building. I was trying to peek around corners, as I often do with new technologies.
Once it wrapped up, I felt a strange combination of things. It was something like one part falling in love; one part kid in a candy store; and two parts watching metaphorical birds flying overhead, harbingers of a massive forest fire on its way.
This post will share some learnings, art winners with highlights, and more.
There were a handful of rules for submissions, but the bolded portions below are the most important to me:
You can use any AI tools or combination of tools that you like, including DALL·E 2 (@OpenAI), Midjourney (@midjourney), Stable Diffusion (@StableDiffusion), Lexica (@LexicaArt,) NightCafe Studio (@NightcafeStudio), etc. You are also allowed to do manual touch-ups and fine-tuning. IMPORTANT: You *must* use Loom (@loom) or other tech to capture your full process.
Selections will be based on quality, creativity, presentation of process, and more. Process is important. People should at least be able to attempt to replicate what you did.
The entire point of the competition was to create this blog post, a simple collection of how-to tutorials that can get anyone up-skilled in a short period of time. Even if you hate everything about AI, here’s why I think it’s worth spending time on this…
The AI genie is out of the bottle. The more true experts I speak with, the clearer this is.
No matter how one feels about AI—and I have plenty of misgivings—the genie cannot be put back in the bottle. This is true for hundreds of sectors and millions of people, including artists. I’ve watched family members work their asses off as artists, I wanted to be a comic book penciler for 10+ years, and I’ve paid traditional artists hundreds of thousands of dollars over the years. It’s a world that is near and dear to my heart.
But… there is no going back to the way things used to be. In fact, there is no compelling way to stall things for long. So the question for me isn’t “How can we stop AI art?,” as I don’t think it’s possible, but rather something like:
“What can artists do to get ahead of the curve?” or “How might proactive artists position themselves to benefit from this?” or “How can artists limit the damage of AI and carve out, or preserve, good niches for themselves?”
There are dozens of other similar questions we could ask, but I think the initial answer is the same for all of them: Learn to use AI tools.
To kickstart your art career, and even if you plan on using more traditional methods… learn to use AI tools.
To turbocharge your own art by streamlining mundane tasks… learn to use AI tools.
To learn the current weaknesses and blindspots of AI tools so you can plan around them… learn to use AI tools.
Most high-calibre artists have nothing to fear, many would-be artists have a lot to look forward to, and all artists have something to learn. If you would spend 10 hours at a weekend course that promised to improve the odds of a good living for the next decade, I think you should commit to spending 10 hours seriously using and scrutinizing the better AI tools. Even if you hate everything this represents, even if it’s an opponent you want to defeat, this is akin to watching tape of an opponent before a boxing match. Whether you want to surf the tsunami, dive through it, or get the hell out of the way, you need to understand the wave. To do that, you need to study it, and the sooner the better. The best defense is sometimes an early offense.
I hope you find this post helpful, and please leave comments with other recommendations. I’m sure I’ll get a few gems and a boatload of screeds, but such is the spice of the Internet.
This blog post is a rough draft and merely a starting point. And if you think the pace of AI development has been fast in the last six months, the next 12 months are going to make your head spin. I think the key is always returning your eyes to the road ahead of you instead of the rear-view mirror.
Just a few notes on structure before we dive in:
We included both “Winners” and “Honorable Mentions,” but all are included because they offered unique value or a unique approach.
I heavily weighed the incorporation of character traits. As I mention again below, it’s easy to make *something* pretty with AI, but to make a very *specific thing* with parameters and characteristics is much harder. For that reason, I tended to reward rougher art with better trait matching vs. polished art that used fewer traits.
Some tweets are no longer live, and we’ve indicated that with a comment.
Special Honorable Mention
I am putting the below Twitter thread by @JPEGHODL first, as the thread itself represents an excellent Midjourney 101 tutorial for people who haven’t used Discord much. This is placed here upfront, as many of the submissions use Midjourney:
Tim’s comments on the below: This fella’s how-to blog post is fantastic. Highly recommend.
Original tweet by @ironclawai: https://twitter.com/ironclawai/status/1604528121360826373
Blog post with all process videos: https://ironclaw.me/2022/12/18/cockpunch-ai-characters-contest/
Tim’s comments on the below: Love this one. If you expand the Loom and watch and pause carefully, you can see the iterative process and also how disobedient AI is with some descriptors/prompts.
Original tweet by @Roof_boy_6ix: https://twitter.com/Roof_boy_6ix/status/1604672157744877568
Process video: https://www.loom.com/share/3b395c838e8e4e098e7a2c633aa88f24
Tim’s comments on the below: Matches the below character well. Good touch-ups at the end, plus use of Photopea.com.
Original tweet by @FPOkenny: https://twitter.com/FPOkenny/status/1605265213808599040
Process video: https://www.youtube.com/watch?v=mA-Xsk-WkbM
Tim’s comments on the below: Great showcase of Midjourney + DALL·E for removing features vs. inpainting, plus shows some common limitations (e.g., hands), though those limitations have been improved recently. Good matching.
Original tweet by @nat_sharpe_: https://twitter.com/nat_sharpe_/status/1605267590670082048
Video process: https://www.loom.com/share/c91e8b09c0c54fb3bedcd1804996db26
Tim’s comments on the below: This process video is amazing. I love the AI + physical art chops, especially starting with absolutely no experience with AI.
Original tweet by @shizudelphia: https://twitter.com/shizudelphia/status/1605316934631051264
Process video: https://www.youtube.com/watch?v=5H3hurVusN4
Tim’s comments on the below: Excellent at incorporating the actual character elements! His faithful inclusion of traits is quite likely the best of the bunch. That was one of the biggest challenges for everyone who competed. Making something pretty with AI is easy, but getting it to make something with well-defined characteristics is much harder.
Original tweet by @PSHoudini: https://twitter.com/PSHoudini/status/1604796676182532096
Process video: https://www.loom.com/share/8558d7597d3e414f8c69e0afb5810182
Tim’s comments on the below: Great explanation of Midjourney and fine tuning with Photoshop. I suspect that, more and more, both artists and writers will use AI to brainstorm rough drafts or starting points for heavy manual polishing.
Original tweet by @blockheim: https://twitter.com/blockheim/status/1605044205906165760
Video process: https://www.loom.com/share/1c6fe71de9b040dc97a8b46090d61941
Tim’s comments on the below: Fun use of traits and incorporated lore (e.g., Dakus), and the process video is also well done.
Original tweet by @WriterInVerse: https://twitter.com/WriterInVerse/status/1605387722033893378
Process video: https://www.loom.com/share/3d5bfe11e4634c889c5ea5546dc7a18c?t=6
Tim’s comments on the below: Thematically strong, rapid iteration, great narration.
Original tweet by @FLitz: https://twitter.com/FLitz/status/1605380881039773696
Process video: https://www.youtube.com/watch?v=jQv5Xr2qJVA
Tim’s notes on the below: This is a good demo of quick Midjourney use, as well as use of variations and upscaling options. Most honorable mentions are honorable mentions because fewer traits were incorporated. Nonetheless, this end product really ended up eye-catching.
Original tweet by @Shakkas_Art: https://twitter.com/Shakkas_Art/status/1605029397710667776?s=20&t=mZHz0kVbGrVT4NMVPR7ACw
Process video: https://www.loom.com/share/2832c4abc7ad4ff0bc5de81959d2d031
Tim’s notes: Good process post, which I appreciate, as I sometimes digest things better via text. This is sometimes easier than freeze-framing the videos.
Original tweet by @BMAotC: https://twitter.com/BMAotC/status/1605157342550573057
Process blog post: https://medium.com/@BMAotC/the-ai-art-competition-beginneth-8d8d7cf70531
Original tweet by @gojeanyn: https://twitter.com/gojeanyn/status/1605397341875277824
Process video: https://www.loom.com/share/1a87f89388104fc791a483cb07f2df10
Tim’s notes on the below: The manual addition of the nose/earring is great.
Original tweet by @moonordust007: https://twitter.com/moonordust007/status/1605375068624019456
Tim’s notes: Excellent process video and shows the refinement of prompts, batching variations, then the finishing in Photoshop. I’m a sucker for Japanophilia.
Original tweet by @FLitz: https://twitter.com/FLitz/status/1605394944075567104
Process video: https://www.youtube.com/watch?v=cgHbP9XfDBQ
Tim’s comments on the below: This one simply made me and Kevin Rose laugh, so I’m including it for fun.
Original tweet by @PiratePorter_: https://twitter.com/PiratePorter_/status/1604349855878459393
Process video: https://www.loom.com/share/998d7fc1d65940cc88d57eda258703c3
Tim notes on the below: This is clever. If you make it hard to see the attributes, you can dodge one challenge. I’ve seen some comic book artists do this with very dark styles that use a lot of black fills (e.g., shadows).
Original tweet by @Whisky_McTavish: https://twitter.com/Whisky_McTavish/status/1604187262253948929 [Note: This tweet is no longer available.]
Tim’s comments on the below: Everyone loves a good steampunk samurai…
Original tweet by @flamesbria02: https://twitter.com/flamesbria02/status/1605162558683746306
Process video: https://www.loom.com/share/27959b7fe1ff42828bae5e43fa301949
Tim’s comments on the below: This one is very good for including detailed gauntlets, which seemed to be one of the hardest parts of the assignment.
Original tweet by @FromInfinityTV: https://twitter.com/FromInfinityTV/status/1604899487088246784
Process video: https://www.youtube.com/watch?v=v-G1cZLMM1Y
Tim’s comments on the below: Great combo of AI and manual tweaking. Too fast to replicate, but it was still worth including.
Original tweet by @Singularifica: https://twitter.com/Singularifica/status/1605090514167943168 [Editor note: This tweet is no longer available.]
Process video: https://www.loom.com/share/ec460d28935b4ac3a93389958332db4e
Tim’s comments on the below: Funny “boomerang” story; great “let me go off menu” experiments. Cityscapes, coliseums, etc.
Original tweet by @DanSchoonmaker: https://twitter.com/DanSchoonmaker/status/1605331136783204355
Process blog post: https://www.schoon.me/blog/cockpunch-ai-art
Screenshot of prompts by @shmichael_eth: https://twitter.com/shmichael_eth/status/1604612837036490752/photo/2
The Tim Ferriss Show is one of the most popular podcasts in the world with more than 900 million downloads. It has been selected for "Best of Apple Podcasts" three times, it is often the #1 interview podcast across all of Apple Podcasts, and it's been ranked #1 out of 400,000+ podcasts on many occasions. To listen to any of the past episodes for free, check out this page.