AI Art

This series is a spinoff of the previous "Lich Princess" series. I wasn't finished with the idea, and testing the same prompt with other models led to XXMix, which gave some intriguing results. Tinkering with the prompt and LoRAs led to some of my favorite images to date.

I spent quite some time editing the resulting images, mostly cleaning up unnecessary artefacts and dust particles, as well as repairing anatomy details like eyes and teeth. I feel that the images look more polished now than I am used to for SDXL images. I also had fun adding some small details here and there. In this image, for example, I added the skull's eyes. It makes me smile ^^

In this serie, I ended up with 129 images that I liked. I broke this down into 39 that were my top favorites, which was way too many. I tried to select only five of these, but it was impossible. After much heartbreak and deliberation I was able to reduce it to these seven images.

Porcelain, ivory and bone. Don't mistake her delicate features for fragility though. She is a master of her craft, and will make you eat your own bones (wow, that sounds more hardcore than I intended now that the words are written).

She reminds me of the actress, Anna Taylor Joy. I feel that she could play a lich very well - which I mean as a compliment.

Lich glamour, if you believe such a thing can exist, is her.

Well, why the hell not? If we humans can become liches, why not fairies? It's an unusual combination, I'll grant you, though personally I feel that the charm is undeniable. Some of the fairy innocence carried over (even if it is very likely a sweet trap).

This shot was incredibly hard to choose over another image in the series. They both had something special, and they both had little fairy wings. In the end it took counsel from the rest of my family to eventually settle on this. I am happy with the choice now, even if I still catch myself thinking to add the other one too :D

I don't know about you, but the magic buildup behind her alone would be enough to make me skedaddle. Add her intense stare to the mix, and you know you should make yourself scarce for what's about to happen.

Here the vampire fangs LoRA had more of an effect than usual, but to good effect. I had to fix the teeth somewhat, but otherwise few edits beyond cleanup were necessary.

Unbeknownst to everyone, Wu Zetian, empress of china during the Zhou dynasty (anno 690-705), was a Lich. Known for her political and cultural achievements, she was an expert at using her necromantic powers to support her reign.

Here the XXMix model really shines in its successful depiction of a regal empress. When I saw this shot, it immediately brought images of a royal chinese court to mind.

This princess is as pretty as she is dead :D

I have recently been reading the Aspect Wars series by Jesse Sprague, which features a heroine that controls death magic, and has a bone hand. It inspired me to create this series to try and capture how I imagine that character.

I have not come entirely close, but I like the result all the same. The text to image process it still too limited to accurately convey meaning. I am sure that this will change in the future as the models get better.

I chose the Nevolution checkpoint because it is great at the digital painting style I was going for, and because for the base prompt I used, it generated the best backgrounds, as this shot illustrates well. The skeleton in the well of light was a unique fluke over the one hundred and eighty images I generated.

Funnily enough, the part "flying flaming skull" in the prompt never yielded any actual flying flaming skulls (it made some hovering ones that were not on fire), but it did contribute substantially to the generation - enough so that I could not remove it.

If you wonder why I have the vampire fangs LoRA in the list, that's mostly because it adds to the gothic style overall, and slightly modifies the mouth shape to good effect.

So yes, this has a hovering purple skull, and one that is on fire but attached to her staff. This is the closest the whole series got to actually fulfilling my prompt "flying flaming skull" :D

This is a mouse that will follow you everywhere, with elegance and simplicity. It boasts masterful precision and ten programmable physical buttons, offers motion assistance and is easy to clean to boot.

This series comes from an idea I had a very long time ago: A mouse with motion assistance. Whenever you move it, it helps you move. The Ziov checkpoint was perfect to bring this idea to life.

I tinkered around a LOT until I found what I wanted, then it was a matter of adjusting the prompt color-wise and do a what I call "seeding runs": Let ComfyUI generate large batches of images (about a hundred), then filter out what looks best.

Small hands, begone! This is a mouse for giants. It is MASSIVE, and it is AWESOME.

• Feel the vibration of the integrated fusion reactor as it spins up.
• Enjoy the assistive wheels anticipating your every movement, making your hand feel as if it moved on a cloud.
• Fully customizable buttons with force feedback - never again will you press a weak button that makes you wonder if you pressed it.
• Add up to 4 Kg of additional weights in the internal storage to get the inertia just right.
• Engage the automatic brake system if you want to do some hand and arm muscle training, with one to ten resistance settings when moving the mouse.
• Last but not least, a super tensile cable with a lifetime guarantee.

Welcome to a new era in mouse lighting! You have the entire rainbow at your command, to be used in one of the forty-two programmable lighting zones. With our Raibow-O-Matic software, paint the colors you wish to use directly onto the mouse in three dimensions!

New with this version is the cable pass-through feature: Connect a second mouse or other devices directly into the back of the mouse, or use it for charging your devices.

Of course all buttons are programmable, including our world-famous side-button innovation adding a third main click button. Make sure you choose the right version - left-handed or right-handed.

You fancy yourself a professional gamer? Then this mouse is a must-have. It is an effortless synthesis of power and style. The pinnacle of elegance in gaming. Behind its sleek aesthetic you will find a powerhouse of functionality and motion-assistive technology.

No less than twelve programmable physical buttons will let you set up any gaminmg scenarios you may need.

As we like to say, you drive - and we provide the wheels.

Don't let this mouse's size fool you: There is a lot of power under the hood. The high wheel torque available through our completely redesigned gearbox and the optimized power output of the miniature fusion reactor make this our most powerful - and comfortable! - mouse to date. It will guide and enhance your slightest movement.

This series of shots is a lucky spinoff of the "Future Tech Suit" series. I wanted to change the suit and hair color, and the result instantly made me think of retro scifi posters. I found a great LoRA to support this style, and I fell in love with the result. At just .6 weight, the "Retro Futurism" LoRA helped adjust the overall color tint and influenced the design of the suit and buildings to give a more retro look.

Possibly the greatest thing about this shot is her smile. For once it does not seem forced or slightly off - it's just a genuine smile, and it makes the shot come alive.

Another shot that really comes alive with her smile, and that has all the retro scifi vibes I was looking for in the series.

I liked the movement in this shot, and her determined look, even if there is a weird recess in her right buttock. One could possibly explain that away with some technical details.

I feel that she could be a space cadet in Star Trek ^^.

This shot is the embodiment of what I see when I think retro scifi. The muted pastel colors, the style of the buildings, the planets... even the shape of her breasts.

This shot took quite a bit of editing to fix its biggest issues. A known effect is that the farther away the face is, the worse the details get. If the issues are too glaring, upscaling the image and running the face detailer does not help. In this shot, the teeth and the right eye were really bad. In fact, there was no right eye, just a blob of skin. I was able to use an eye from another shot in the series, and the teeth I could partially repaint manually. While it still looks strange from up close (I am reaching the limit of my Photoshop skills), it's good enough not to notice something amiss right away.

This series was inspired by a manga image I came across on the Civtai website. Working on the prompt, the results quickly went way beyond what I thought was possible. This is now one of my top favorite series. This was also largely possible due to the DucHaiten checkpoint - only two LoRAs were used for the whole series.

Believe it or not, I got this hair style to work the way I wanted it by using "blonde super saiyan hair" in my prompt. Realizing that this kind of prompt can work opened quite a few possibilities.

This was also the hardest series to date for which to choose the five images to showcase here. There are so many great shots that I ended up with a final selection of thirteen favorite images. Selecting only five among those was almost physically painful.

This shot has a decidedly supherhero vibe to it, and made me think of Captain Marvel for some reason. This made it stand out from the rest of the shots in the series, and made me include it in the final selection even though it is not the most impressive.

I have always admired people who wear their clothes with such confidence that they can wear virtually anything and look fabulous in it. This is one of those cases, only that the bodxy suit is fabulous to begin with. It's nuts.

I never would have thought that an image generation AI would be able to create such a great rendition of a chrome-like surface. It continues to blow my mind. Also, using Photoshop's "Bicubic Sharper" image resizing mode really made the edges stand out without looking overly sharpened.

Organic, sleek flowing chrome. Sign me up, I want to leisurely stroll through the neighborhood with the same kind of suit.

Note the seed on this one: 42. It seems fitting that Douglas Adams' famous number should have generated this great shot!

This series was loosely inspired by a mystery book I have started to read. I knew I wanted a detective lady that likes tea and cookies, with a rugged nerd kind of look. It took quite a few tries (~180 images), but I really like how it turned out.

The main issue here was the hands again. Mostly the wrong amount of fingers, but also quite a fair share of disconnected hands and warped fingers. Even in this selection you will notice these issues, although I feel that they do not disturb the overall impression of the images.

It still amazes me how the models can get so many things right, from the face to the clothing, textures and more, and be so bad at fingers. Then again, having done my fair share of drawing in a family of artists, I know that fingers are quite a challenge. So maybe it's not that surprising after all!

This is an earlier image in the series (#48 to be exact) which was one of those flukes that turned out really good. It has this touch of magic that I was not looking for, but which worked well here. The weird lightbulb, posters and her pendant add a layer of mysticism.

Those smoking mini-cookies look scrumptious ^^

I really like the light in this shot, and the open sky in the background. It has an open feel to it that the others do not have.

Also, yup, that's six fingers. Funny how you don't see it at first, then wonder what's wrong until you realize that hand is strange-looking.

This scene was kind of an oddity. There we six checkpoints that generated recognizably the same place in their own style with the same seed. In some, even the placement of the ducks was eerily similar. This shot here was done with "Halcyon", and the next was made with "JibMix", and you can see how similar they are.

From what I learned so far, this makes me think that both checkpoints share some of the data that they were trained on (maybe there's a set of water mill photos that several models used?). Since JibMix is a merge checkpoint, it's possible that parts of Halcyon, which is a trained checkpoint, went into the mix. In any case, it's very interesting to compare the two scenes with Halcyon's soft edges and JibMix' photorealistic style.

It's been several nights now that these saucers have grown little stone pedestals and are just sitting there looking pretty.

This series started out with the idea of showing a group of flying saucers hovering above Stonehenge. The prompt slowly evolved as I saw the possibilities and played around with LoRAs. I quite like the softness of the grass and the retro scifi look of the saucers.

This is a very early shot in the series when I was just starting out with the prompt. Even though it is not as detailed as the others, it has this little something that made me choose it among the eight images from my final selection.

Fun fact: The saucers can still detach from the towers to fly around. Very practical to fetch that sunday morning croissant or to go grocery shopping on the second moon (when there are no lightning storms).

This was one of my wilder mixes with 11 LoRAs working together. It was really hard to choose a shot, because a lot of them were great. In the end this one prevailed because of the nicely winding river and the big planet in the sky.

Glowing eyes? Check. Gothic baroque outfit? Check. Creepy & intense stare? Check. Consider me gone with the wind, as fast as my legs can carry me. I'm just hoping that I can run faster than a vampire bat can fly (spoiler: I can't ^^).

Oh, and if you're wondering what vampires are actually like, I invite you to read my research paper, "On Vampires".

This series started out as a mage in a greek temple, and somehow ended up looking decidedly vampiric. Using just a little bit of the "Fangs" LoRA brought the vampire out fully.

Some great details here are the jacket's collar and the baroque jewelry. Even though no fangs are visible, the half-open lips and tilted head manage to convey the imminent danger pretty well, I believe.

Yup, nope. Ain't no such thing as an innocent vampire.

Unless you fancy yourself a philosopher and want to delve into the nature of vampires and choice vs instinct, magical compulsions and how much free will a vampire retains when the need to feed makes itself felt.

Yup, nope. No such thing.

Well, that's an invitation that seems innocuous enough, and the company does seem charming. The candlelight sure is a romantic touch too, and the carpeted dance floor is exquisite. It would have been even better with some mirrors, but this will do nicely anyway. My lady, lead the way.

This shot shows how close I had gotten to having a full-blown vampire without even using the "Fangs" LoRA.

This series resulted from the very first images I generated with ComfyUI. I came back to it with new knowledge to give it the attention it deserved. I was going for traditional to cheesy fantasy, and am very happy with the results.

In this shot I mixed in some of the Franz Frazetta (Wikipedia) LoRA (at 30%), which directly changed her robe to a bikini, and added musculature to end up with this chiseled body - which matches the archetype of Franz' barbarian women. Funnily enough it also added elf ears ^^

This was a very long-running series where I tested and learned a lot about ComfyUI and AI art generation in general. It is one of the shots that crystallized after 200+ generations across multiple checkpoints. The combination of tattoos, jewelry and hair style is what particularly stands out in the series.

Once I have a good prompt and LoRA mix going, I like switching out the checkpoint without changing any other settings to see what comes out. In this case it was a really good surprise as the Albedo Base checkpoint generated some very nice shots. This one I like because it has what I feel are some definite Mayan vibes.

I did not expect to be transported to asia when I switched to the Crystal Clear checkpoint, but was very happy with the results. I generated quite a few of these that I love, but chose this one because it is in portrait format like the others.

For this I switched to the Nevolution checkpoint. It generated some really good shots without changing any settings. I especially liked Nevolution's interpretation of the stone monolith in the prompt, which was converted to these two figures embedded in the pillars on each side.

This was a switch to the Cheyenne checkpoint which also generated some very interesting shots. This one stood out not only because of the statues on each side, but also of the regal demeanor and distinctive jewelry that she wears. One of my all-time favorites.

Because she wields two flaming swords, get it?

...seriously, don't. Yes, it is very small, but that's not the point. Well, it does have a pointy edge - okay, nevermind. Anyways, if you mention the dagger's size in any way, prepare to be crispified. This flame atronach is working through some issues.

Prepare to be smitten, whoever may need smiting! Which is not very many people or entities in this barren, molten volcano caldera. Clearly not the best venue to smite anything, really. Still, the pose is very impressive.

This shot has something biblical in its pose, like Saint Michael smiting a demon with his lance. I know that scene intimately because I when I was little, my mother drew a graphite sketch of it. The canvas seemed very large to me back then, and the scene impressed me a lot.

This shot is why I love generating AI art. It's a one in one thousand fluke that generates something truly unique and amazing, and I count this shot among them.

The Cheyenne checkpoint really excels at creating original color and style compositions, as this shot shows. The only thing I had to fix were the eyes. They were really bad, and even running the facefixer did not help. I ended up reusing the eyes of another shot in the same series, which helped a lot to get both the colors and style right.

The face was also a stroke of luck. In this shot, I mixed in a bit of the "Yennefer" LoRA that adds the likeness of the Yennefer character from the video game "The Witcher 3". This is one of the very few shots where that gave good results.

In this shot I had to do some more extensive fixing of her chin. It looked as if there should have been some cyberware there, but it ended up just looking botched. This is where I cannot praise Photoshop's "Healing Bush" tool enough, as it allowed me to fix the skin texture while keeping the overall lighting intact.

Anyone else think that she looks a bit like Zoe Saldana? It is pure chance, but it looks really good :)

The "Neon Noir" LoRA made this shot so dramatic. For the other shots I felt it to be too dark, but in this one it really shines (see what I did there?)

This is a spinoff of the "Cyber Viking Queen" series after tinkering with the LoRAs to switch the color scheme somewhat. The result is something quite close to the Cyberpunk 2077 palette and overall styling.

This was one of those shots that stands out from the rest. The monochrome palette was a total fluke compared to the other color-heavy shots, which instantly made it interesting. Also the fact that she is perfectly centered with the portal behind her made me select it for the gallery.

Sorry for the title. I know it's bad. Really bad.

Many say that the crown of molten feathers never shone as bright as it does with her. On the day of the coronation melding, the circuitry of the whole megapolis brightened so much that it made the night as bright as day. There are rumours that her thoughts and even her spirit infuse the net, as was long fabled possible.

Those eyes have been known to make seasoned warriors weep. Don't be fooled by their innocence and her youthful demeanor: This queen is over four hundred years old.

Her cyberware slowly grows and consumes her, but her humanity is etched into her cybernetic components. If she looks at you critically like this, know that you are not judged: you are measured for your entire existence, and she understands and shares the path that you have chosen to walk.

She has a nice smile, but don't let that fool you. Freshwater mermaids are no pushovers, and have a very competitive battle culture.

This series was me wanting to do something more photorealistic, but still keeping the fantasy aspects. The idea of freshwater mermaids living in the jungle was something I had been thinking about for a while.

For some reason, I did not get her to wear full body armor coverage. No matter the prompt and negative prompts, I always ended up with these bits of armor that only protect half of the wearer, typical for male-centric fantasies. Even though they are living in a hot area, I cannot imagine their male counterparts wearing as little as this. I will come back to this series later and try again to get them to wear some proper armor.

That lance may look ornamental, but as many of her foes cannot disprove anymore, it is effective.

Hair is a big deal for the freshwater mermaid tribes. Shilli here is known far and wide not only for her massive mane, but also for being one of the first to complete the mermage trials (hence the body tattoos). Beautiful? Yes. Deadly? Very much so.

...one step further, and you'll get to see this blade from closer up than you'd fancy.

I know I would be wary of an armored lady advancing on me with that look, however stylish she may be.