What’s your background and how did you get into AI art?
I displayed some talent for drawing at an early age and was always creative. I took private lessons with professional artists when I was young and was the “class artist” in school. I had to choose between going to art school or majoring in Comp Sci for college and chose the more practical and financially stable path.
But I couldn’t stay away from visual arts and craft. Even my software engineering career has always been focused on UI and front end, and I was always doing something artistic in my free time from jewelry making to furniture refinishing to photography and digital art. I never had enough time to focus on art as a career though.
Though I experimented with style transfer as early as 2016, I really found AI art through NFTs and NFTs from getting into crypto in 2021. A friend I made in my first NFT project community pinged me one night about this “text to image” AI NFT project called Eponym that was minting out fast. He knew I would be interested and he was right. “Just type ANYTHING and it will make an abstract picture of it!” I was immediately smitten. I didn’t make something worthy fast enough to mint that first generation, but I became heavily invested in that project and for the Gen 2 and PFP spin-off.
With that first mint, I discovered I could keep the generator open in a browser tab and keep making images even if I couldn’t mint them. I spent weeks constantly generating things in every spare minute just to see if I could suss out consistent prompts before eventually the 3 tabs I had open crashed. I kind of went through withdrawal then, until Gen 2 was released that could be used to generate any time. Eventually the project and community kind of slowly splintered because of some internal drama.
But in the meantime an amazing piece of AI work like nothing I’d seen before popped in my Twitter feed made by Mark Kelly. I told him to mint it because I had to have it. He did, and that purchase became a very eventful connection for me. Mark turned out to be an incredibly kind and helpful person and shared with me how he had made the piece I was enamored with and even pointed me in the direction of the software he’d used: Visions of Chaos. I realized that I could run my own open-sourced AI instances and have so much more control than I had with the proprietary Eponym AI, and it could be something that I could modify and perfect with post processing, too.
Fortunately my husband had a gaming PC with the minimum requirements and so I basically hijacked it to install VoC and began my in-depth AI art explorations, starting with the algo Mark used: ruDalle. After a month or two I made some work I was pretty proud of as collages of many generations of experimentation. I put my first collection on Foundation in February of 2022 and after selling it and a second collection out, ended up getting my own even beefier PC. I learned to use Latent Diffusion (LAION), and Disco Diffusion, then got in the beta for DALL-E and Stable Diffusion after that. Stable Diffusion and all the models and tools in Automatic1111 have become my favorites. I have only recently started using Midjourney as well, but my work now is usually passed through more than one algo before it’s finished.
Do you have a specific project you’re currently working on? What is it?
I have been invited to a couple of projects and am working on pieces for those, but no large-scale projects right now.
I do have a project running that I have committed to for a year that I call my “Flower of the Month Club” where I release a floral open edition for an entire month. I am known for my floral and Art Nouveau inspired work, so it’s a chance to share some pieces for a relatively inexpensive price and is an experiment in the vein of Patreon. I am about to wrap up the 6th edition and am thinking of how to reward those wonderful folks who have collected half a year’s worth so far, so that’s probably my next project to focus on.
What drives you to create?
From the deep psychological perspective, I suppose my drive to create is to answer my need to find some pattern in a seemingly random world, put a little order around the chaos of my life, or find the treasure of details that other people ignore.
I was always doing something creative in my “free time” even during my challenging adult life of working a demanding career and raising a child. Creating was an anchor in that constant flow of responsibilities that served the needs of others and constantly pushed me away from my own desires. Creating something I could look at and touch or wear made me feel like I took back some agency to directing my time and added beauty to my life where it was missing.
In the last few years I had some especially difficult times and creating art became a refuge and maybe even lifeline. It is something I have to do daily to keep from letting accumulating depression and anxiety disable me. Creating has been something that makes getting out of bed in the morning worth it. And using AI has made it much easier to make the kind of intricately detailed maximalism that I can lose myself in and forget about my own troubles for a while.
What does your workflow look like?
My workflow has evolved as I’ve built a cache of hundreds of thousands of generations through learning different algos and tools and I have started “reusing” results as inits or blends.
I have two distinct paths that I go down that result in finished work: the “joy of discovery” path and the “firm intent” path.
For the “joy of discovery”, it’s about learning and experimenting. There is an ever-flowing stream of new AI algos, tools, models and versions to keep up with and during the learning process I will inevitably create pieces that resonate with me either just in the look or emotion or that inspire a story when I see them, even though they were not things I set out with intention to make.
For “firm intent”, it’s about having a vision in mind that I want to create and using the experience gained from the discovery path to be able to render that vision.
Of course my workflow tends to be variable when working on the discovery path, and more static with firm intent.
Most recently my technical workflow is finding two images from my previous work or cache of generations and blending them in Midjourney. Then remixing that blend with positive and negative multi-prompts of terms that are not about the content of the image, but quality and style to find out what MJ “sees” in the image. I love the variations that it can produce starting from the same base, and using my blends as the base gives it colorways, subjects and styles that are already closer to my interests than raw prompting.
After many iterations, I may find a variation that resonates, but MJ’s results are often too “soft” for my tastes and no matter what I try still have an “MJ look”, so I am never satisfied with the raw results.
So I use the candidate as an init image in Stable Diffusion with my favorite “refining” prompt/negative prompt customized for the subject. I iterate with permutations of combinations of my favorite models, samplers and seeds and at different CFG scales to dial in the styles and details I’m looking for.
I also enlarge the piece to 2048px on the largest dimension during the SD processing, which helps to pack in as many details as possible.
Sometimes I find a single result that works for me, but often end up with 3-5 variations I like from different seeds or samplers. I will put them all in Photoshop as layers and use masking to create a composite with the best features.
I spend a good bit of time in Photoshop with every piece correcting things that look wrong to me with the liquify tool, or the healing and cloning tools, and sometimes painting in new details. Occasionally, I will replace larger areas with context aware fill (or the new generative fill) until I’m happy with the piece, but that is rare since if it’s a large enough area, I’ve probably got a variation to collage there.
I will then make a composite layer and use Topaz Sharpen on it to refine the details further. Sometimes I will mask that layer and bring back in softness if the sharpening is too much in some areas.
After finishing the details, I always spend a lot of time on the color, using adjustment layers for color balance, hue/saturation, vibrance, curves and selective color to color correct or move things toward a more cohesive palette (I like quadratic or split-compliment palettes the most).
Once I am done with all the edits, I use Topaz Gigapixel to upscale to at least 3000px on the largest side which is a good size for social media sharing. Depending on what model I use for that, I might need to run the enlarged piece through Topaz Sharpen again.
If I am minting it, then I will do another upscale to at least 8000px, since this is high enough resolution to print decently at 24in x 24in, if a collector chose to print it. Sometimes this also requires some extra sharpening or even touch-up edits in Photoshop again to ensure the best quality.
Regardless of path (discovery or intent), my steps in Photoshop and after are the same, but the path can change what I do with the AI tools significantly.
What is your favourite prompt when creating art?
I use different prompts for Midjourney and Stable Diffusion and in most cases the prompts are more about refining instead of generating content (remixing blends in MJ or “cleaning up” MJ inits in SD), so sharing them wouldn’t be that useful to most.
But I can share a bit of process that I’ve been experimenting with recently in Midjourney with blending. Most of my previous work is very maximalist and in some cases heavily abstract, so I’m interested in seeing what MJ will “figure out” the image is “about”. I don’t do photorealistic work, so my suggestions here may not be effective for that.
Once I have made a blend, I have discovered that remixing with simple multi-prompts and negative prompts can yield some very interesting results. I will start experimenting with just 1-3 of these at a time and add or swap them until I get results I like:
And to counter the “MJ look” will add this negative prompt:
That usually has a big impact. I will sometimes bump the weight to
-2 or even
-3 if I have enough positive prompts to offset (the sum of weights must be
0 or more).
I also find that using some “medium” prompts with those in the first list can add interesting results:
ink and watercolor::
Or sometimes I will include some of my preferred art styles such as
The key to all this is that I’m only using quality or stylistic terms for prompts and nothing related to content or subject matter.
I will often also play with these prompts with just a single image prompt after doing a blend since I can reuse the shortened url that MJ provides.
It is all tons of experimentation and building on images over and over, but it is a process I really enjoy.
Can you tell me something about the AIIA DAO you’ve founded?
The AIIA DAO is a project-focused group that was initially formed with members from the more movement-focused AIIA (AI Infused Art) group that I also helped to found. The group was formed to operate like a DAO where the members would be leading in suggesting and implementing projects through majority consensus. It was intended to initially help us market work as a collective which could both aid in member sales and fund a treasury to do more projects, as well as eventually provide services to the AI artist community such setting up an art vault and sharing knowledge, techniques and learnings with publications or spaces in the name of promoting AI art acceptance.
We are still a relatively new group just a few months old, but we are growing as members nominate others for membership and the DAO votes to bring folks in. And as we grow, our mission and goals are changing a bit, as things do in a democratic group, to serve the wishes of the members.
Ultimately I wanted to be a part of a group that has some ambitions to work as a collective, support each other and give each member in the community a voice. I did my best to make it happen, setting up our Discord, Twitter and Manifold accounts and financially bootstrapping things like minting membership tokens and getting an ENS name. I also owe a lot of gratitude to my co-founder, @knylx_art, whose shared vision and contributions of time, effort, infrastructure and morale I could not have done it all without.
How do you imagine AI (art) will be impacting society in the near future?
I think that the biggest impact as disruption has really already happened for AI art. We went through a phase of shock for traditional and commercial artists which resulted in a backlash, but I think we are starting to move into an acceptance phase now, however reluctant for some that may be. It’s no longer a novelty.
While whether models have been “ethically sourced” is still up for debate (both legal and by public opinion), the genie is out of the bottle and it will not be put back in.
People are beginning to get used to the idea that AI as a valid tool for artists to use exclusively or as one in a more traditional process, and we are moving toward it being as (not) controversial or noteworthy as using something like Photoshop.
Things are progressing at such a rapid pace that what seemed impossible (”normal” faces? “normal” hands? super photorealistic images?) a few months ago is almost taken for granted now.
We are on the verge of high quality video production, image manipulation by just dragging around control points and other innovations that will enable creative people to produce even more impressive work efficiently.
Within a year from now, I think using AI both in art and in general will be much more mainstream and powerful than we can accurately guess right now, and while that’s exciting, I am also a bit concerned about things like model collapse (training a model on data that AI produced and eventually exaggerating all the flaws in it), or how AI will influence the cognitive development of young children who are exposed to it as a learning tool if we can’t eliminate things like hallucinations, or even having AGI emerge and not having alignment in effect which poses a threat to our very existence.
As someone who has followed new technology with enthusiasm since the 1980s, this is the first time in my life that I can’t confidently picture what things will be like within 5 years, a year or even 6 months from now and that is both exhilarating and scary!
Who is your favourite artist?
My favorite artists and those that influenced me from an early age are Rene Magritte with his clean surrealism that wasn’t full of deformities, M.C. Escher with his tessellations and maximalist patterns and Walter Inglis Anderson, a local artist to where I grew up who made intricate woodblock print and watercolor paintings that were often surreal depictions of nature as patterns.
From my teenage years and on I fell in love with Art Nouveau and Art Deco and enjoyed the work of artists such as Elisabeth Sonrel, Walter Crane, William Morris, Alphonse Mucha, Tamara de Lempicka, and Sandro Botticelli who all had extremely detailed, organic and ornate work. I also love the bold lines, colors and nature depictions of Ukiyo-e artists Katsushika Hokusai, Utamaro Kitagawa, and Utagawa Kuniyoshi.
All of these plus illustrated fairy tale and Greek Mythology books that I poured over as a child have had great influence on my art style and preferences.