Is it Time to Replace Your Artists with Artificial Intelligence?

[zanegun08]

The ability to generate creative designs on demand will accelerate design and production cycles and will make it easier for brands to create new products without having to invest in expensive talent or equipment. This is a revolutionary time for designers who can now supercharge their creative process with artificial intelligence.

Artists are the ones who bring our ideas and concepts to life, they are the ones who take the blank canvas and turn it into a masterpiece.  However, with AI, you can now generate art without a traditional artist. This means that you can create designs with little artistic skills or talent by simply writing descriptions of what you would like to create.  AI art generation is becoming a reality with the public release of platforms to create art by text. You can generate these things at a rapid pace which means that you don't have to wait for your artists to finish their work on time.

This article explores the benefits of using AI in your creative process, as well as some potential drawbacks and concerns.  With the public release of AI art generation, can you replace your artists?

Up until this point, this article was mostly generated by AI, just by entering a prompt and some keywords.  This is also how programs such as Midjourney, Dall-E 2, Stable Diffusion and others operate to create images.  Simply type in a phrase and description of what you would like to generate, the style, colors, mood, rendering style, and even reference popular artists to new designs.

Midjourney, which has a free trial and operates through discord, is the easiest to get started.  There are many tutorials online and on youtube which will help you get started.  There is a learning curve to generate t-shirt ready artwork.  For the purposes of generating apparel designs, you can modify your aspect ratio to fit the canvas area to get a design that will fill out a shirt, but getting something without hard edges can be challenging.

Since Halloween is coming up, let's create some design concepts for the holiday.  Using the prompts below to rapidly generate some concepts will get you this.

“haunted house, full moon, black background, Dan Mumford style, intricate design, low color count --testp --ar 1:2”



“jack o'lantern, black background, ed roth style --testp --ar 1:2”



“satanic ritual, fantasy style, photorealistic, hyper detailed --testp --ar 1:2”



I like the jack o'lantern design, so I took one of those into photoshop to make some quick edits, colorize and halftone, and then back into illustrator to add some text and mockup on a shirt.  Without the use of AI artwork, I wouldn't be able to create this style of design as it's outside my skillset.  But leveraging the image creation opens up the ability to create designs that would not be possible.

Here is the final design ready to go as a two color with halftones, this would print really nice with discharge ink.



The reception of AI art has been controversial, with many artist calling it soulless. But just as there was a time when photoshop and illustrator was ill-received, the people who embraced the technology were able to excel in the long term.  Photoshop has been adding more AI assisted tools, and if you've used content aware fill, you've already been using one.

So will you be replacing your artist with AI anytime soon?  I think we are still years off, as you still need to be creative enough to write prompts, and have the skillset to finalize a design, and an eye for composition, but I see it as a way to supercharge your artists with what they couldn't do before, and quickly.

For more, and much better examples of what can be created with AI, check out the Midjourney showcase https://www.midjourney.com/showcase/ which includes the prompts used to create the images.

-------------------------------------------------------

I reached out to impression magazine to see if I could write an article for them but never heard back, so I'm posting here instead.

https://www.midjourney.com/

https://openai.com/dall-e-2/

in the prompts, --testp is for the beta version of midjourney which makes supposedly better images

[blue moon]

Awesome write up!
Thank you!
Pierre

[rusty]

This is great. Awesome write up and I i look forward to diving into this tonight.

[ericheartsu]

Zane this is incredible. thank you for posting this.

[sunday]

Very spooky, 'tis the season!

One thing to watch out for is the copyright/who owns what's made with these AI programs. Always gotta watch out for the T&C!

From Midjourney's TOS (https://midjourney.gitbook.io/docs/terms-of-service#4.-copyright-and-trademark):

"By using the Services, you grant to Midjourney, its successors, and assigns a perpetual, worldwide, non-exclusive, sublicensable no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute text, and image prompts you input into the Services, or Assets produced by the service at your direction. This license survives termination of this Agreement by any party, for any reason."

[zanegun08]

One thing to watch out for is the copyright/who owns what's made with these AI programs. Always gotta watch out for the T&C!

I don't see this as an issue, t-shirt art has a short lifespan and there are literally millions of images produced, so there is a slim chance of having any issues, as well as the open community aspect can be used to improve your prompt writing as well as get inspiration for what is possible.

Here are two images I created with Midjourney and had printed with DTG printing.

The first I created the girls face with this prompt "face photo, samurai girl, intricate detailed ivory head dress, looks likeZendaya Maree Stoermer Coleman, moody lighting, red and blue, intricate detailed ivory design, bioluminescent colors, back lit black background, 4k, octane render, epic composition, golden ratio" which created this image https://mj-gallery.com/8150c773-fc71-4031-a7d1-51fcee9daec6/grid_0.png

I then created the background with "chinese scroll writing, black on white, bold text" which gave me this https://mj-gallery.com/991dc25c-7ec3-4a89-93c7-fb5b5f5939be/grid_0.png

From there I cut out the girl in a way without hard edges, and made it more symmetrical, as well as I ran it through another AI https://huggingface.co/spaces/NotFungibleIO/GFPGAN which helps clean up faces as well as AI upscaling.

I then created with this prompt "reflective metallic gold sheet" https://mj-gallery.com/e5c62c96-c46b-4fd2-90cb-ee4a40bcbd64/grid_0.png which I superimposed into the background,

And finally I added the streetfighter logo as it reminded me of streetfighter and I wanted to make it more of a "real shirt design"

Here is the final result that I had printed via a DTG printer as a one off.





The second design, I used the prompt "lil wayne as baby in onesie" and got lucky with a flat background that was easy to cut away, here was the result https://mj-gallery.com/85a2102d-31af-4f38-ba82-e77d3e936ca5/grid_0.png

I then got a high resolution photo of Lil Waynes face to add back in the face tattoos.  This was prior to knowing about the GFPGAN so the eyes are off, but good enough for me on this one.

I then generated some "baby pink chenille fabric" which I typeset some text and superimposed the fabric in the background. 

Since I created this design, Midjourney has been upgraded to make perfect "seamless" images that can be tiled easily.  Here is an example with the prompt "alphonse mucha style painting, octane render, hyperrealism, seamless pattern, flat background light" https://mj-gallery.com/314cd96e-e42d-49dd-973c-6fb8b5558430/grid_0.png





A couple of tips, you can use midjourney without paying on a trial basis, if you open in a private window and just make up a discord name (you won't need to login) you can just close that after your trial runs out and re-do it in a new private window with another trial.  I did end up paying for a month ($30) though as it's good entertainment and I feel will be usable in the future and I can just build that $30 into the price of artwork easily.

Second, watch youtube videos or read some blogs about prompt writing, there are a lot of keywords you can add that will drastically change the image, here is a google doc with examples of the same prompts and adding an artist as a reference https://docs.google.com/spreadsheets/d/1h6H2CqjLdZMbLjlz6EHwemfO4fkIzAfWtjRAMSa2KHE/edit#gid=0

Lastly, share here if you come up with anything or use this for any projects, this could really benefit the right niche of printers who do band shirts, have DTG printing or hybrid printing where color count isn't such a factor, but also for rapid prototyping and then sending off to be vectorized as a more simplified design would also be very beneficial to someone who doesn't have many hours to make artwork.

[rusty]

Thanks for the follow up write up. This looks really promising, and I agree copyright isnt much of an issue since this would be used more often for small runs or people who don't have art and don't want pay alot.  Same as using clip art from all the usual sources.

[sunday]

A couple of tips, you can use midjourney without paying on a trial basis, if you open in a private window and just make up a discord name (you won't need to login) you can just close that after your trial runs out and re-do it in a new private window with another trial.

https://midjourney.gitbook.io/docs/terms-of-service#exception-1-non-paid-members-license-terms

"If you are not a Paid Member, Midjourney grants you a license to the Assets under the Creative Commons Noncommercial 4.0 Attribution International License (the “Asset License”)."

https://creativecommons.org/licenses/by-nc/4.0/legalcode

If someone's just trying it out, then sure, create a throwaway account and give it a go. But if this art is going to be used for a commercial purpose (like making band shirts), then what you're suggesting violates the Creative Commons license in question.

[GoWestRob]

I've messed around with the DALL-E one a bit after seeing a funny bit about it on John Oliver.  It's really impressive with what it can do, but still the results are pretty shitty for production artwork.   Although maybe I need better keywords.  In the past we have used services like clipart.com to come up with something quick when we need a vector 'flying bat' or something simple.  So far I see DALL-E being more like a more customizable version of clipart.  It's a tool you can use to get you a step in the right direction, but of course you'll need a skilled artist to make something amazing out of it.  I don't think it'll 'replace' artists anytime soon, for now it's similar to outsourcing vectorizing to an indian vectorizing service instead of tracing it in house.  The AI can get some of the initial legwork done and then a pro can take it from there.  It's definitely interesting to watch it develop.

[Dottonedan]

2nd time I’ve seen this pop up on the radar in over a few months.  Another artist friend of mine who used to be a member here years ago, turned me onto it. I’m surprised that he was super stoked about it because he’s more of a traditional Illustrator and very anti-employers taking advantage of artist. So he must besting this as a much easier way to generate art income.


I checked it out myself. Did the name thing and tried one (first time) and I am sure I didn’t invest much time needed to really give it a good go.  I’m sure it will be another tool for artist to use and for new shop owners needing to do something in art, it may help greatly.

[Sbrem]

And I like to think I'm up to date, LOL! This is very intriguing to me, this is the first I've seen of this. Thanks for the post, I'll be digging in starting tonight. It's not like I go out on Friday nights anymore...

Steve

[3Deep]

Going to have try this out, as a half ass artist/clipart jockey etc, this could really speed up things I do for customers and get there ideas faster on paper or screen