AI Prompts

[DonShip]

I'm trying hard to create a female elf that looks with this picture's style using Krea AI, but can't seem to get close.
Anyone have any prompt ideas? Or am I using the wrong AI image generator?

[sneakly]

Why not post some of the prompts and see if anyone has ideas? I bet there is an author chat somewhere that discusses AI generating book covers that might have some interesting suggestions.

[Mr. X]

Its not the prompts its the checkpoint and filters you use. Those higher end pics are multiple filters, face enhancements etc. For the stock stuff you get the same elf girl over and over.

[DonShip]

Its not the prompts its the checkpoint and filters you use. Those higher end pics are multiple filters, face enhancements etc. For the stock stuff you get the same elf girl over and over.

Damn

[Mr. X]

If you email DanO at Superheroinecentral.com he knows a lot more about this than me. He does some excellent work.
https://x.com/DanOSHC/status/1849987716479328285

Did you look at civitai for anyone doing a lora or checkpoint for they type you want?
https://civitai.com/
https://civitai.com/models/775455/short ... -girl-pony
For example.

[Bladebur]

I think you can get a similar style with ZavychromaXL (https://civitai.com/models/119229?modelVersionId=563988) if you ask for concept art in a matte painting or comic book style. The following one's prompt was "concept art, comic book, beautiful, close-up of a sophisticated sexy woman, classy dress, long and messy hair, elf, jewelry. digital artwork, illustrative, matte painting" with "photo, photorealistic" as negative.

Not sure if it is what you're looking for.

elfw2.png (1.59 MiB) Viewed 13063 times

The original images you posted look to me like heavily edited or iterated SD1.5 images. Getting a consistent style out of an AI is very hard.

[DonShip]

That's pretty great. Thanks.

I might have to move to a different AI, since I haven't found any "models" on KreaAI.

[Mr. X]

KreaAI is a wrapper for stable diffusion

[argento]

KreaAI is excellent for improving an image, but it has some shortcomings. For instance, it struggles with facial expressions and deformed fingers, which is something I've noticed in other AI as well. I don't use 100% of the AI, and I keep the original face, hands and feet. I don't know how many images you want to make, but if it's a lot, use a female elf model from Daz 3D to make exactly what you want. Use the renders you make with the AI proposed by 'Bladebur' for comic style painting.

[Mr. X]

It also hates masks or anything on the mouth.

[Bladebur]

That's pretty great. Thanks.

I might have to move to a different AI, since I haven't found any "models" on KreaAI.

Krea has a number of Stable Diffusion models available (if you enter real time mode, in the lower bar you have a widget to choose your model). The 'Concept' model seems similar to the one I used above. It's a shame they don't reveal which models they actually use.

Krea also supports Flux. It's an awesome model, although it's hard to get a specific style with it. I got the following elf with in Krea Flux with the following prompt: "Comic book cover art of a beautiful, modern and sophisticated elf woman with long white hair. The illustration is painted on oils, a matte picture with strong lines, in a realistic western style. The woman is wearing a classy dress with jewelry and is smiling to the camera in a suggestive pose. The background is an abstract mix of colors and the lighting is extreme. It's an interior scene, set at dusk."

Perhaps you can find something you like by trying similar prompts. The problem with Flux is usually getting something that is actually too detailed or photo-like and doesn't look like a painting.

comic_book_cover_art_of_a_beautiful_modern_and_sophisticated_elf_woman_with_long_white_hair_the_ill_adflsg71gdkada3i26u0_1.png (1.69 MiB) Viewed 12683 times

[DonShip]

It also hates masks or anything on the mouth.

I find it often deforms nipples, too.

[DonShip]

That's pretty great. Thanks.

I might have to move to a different AI, since I haven't found any "models" on KreaAI.

Krea has a number of Stable Diffusion models available (if you enter real time mode, in the lower bar you have a widget to choose your model). The 'Concept' model seems similar to the one I used above. It's a shame they don't reveal which models they actually use.

Krea also supports Flux. It's an awesome model, although it's hard to get a specific style with it. I got the following elf with in Krea Flux with the following prompt: "Comic book cover art of a beautiful, modern and sophisticated elf woman with long white hair. The illustration is painted on oils, a matte picture with strong lines, in a realistic western style. The woman is wearing a classy dress with jewelry and is smiling to the camera in a suggestive pose. The background is an abstract mix of colors and the lighting is extreme. It's an interior scene, set at dusk."

Perhaps you can find something you like by trying similar prompts. The problem with Flux is usually getting something that is actually too detailed or photo-like and doesn't look like a painting.

comic_book_cover_art_of_a_beautiful_modern_and_sophisticated_elf_woman_with_long_white_hair_the_ill_adflsg71gdkada3i26u0_1.png

Thanks. That is very much the kind of images I'm trying to create.

[superheroinelinks]

Many don't like AI but I disagree. I Love it

[argento]

elfa 1.png (1.41 MiB) Viewed 12402 times

elfa 2.png (1.36 MiB) Viewed 12402 times

elfa 3.png (1.22 MiB) Viewed 12402 times

elfa 4.png (1.27 MiB) Viewed 12402 times

elfa 5.png (1.72 MiB) Viewed 12402 times

elfa 6.png (1.5 MiB) Viewed 12402 times

Sorry for the delay. I had some personal issues to deal with. I still believe a DAZ model would be a better fit. The AI lacks sensitivity.

[BlackOut]

The only thing I use Krea for currently is to enhance videos, which is especially useful when I'm putting together multiple clips since it does a good job of hiding any color/tone variations between the different clips.

For example. This is two clips joined into one.

ezgif-2-11db665c2e.gif (7.49 MiB) Viewed 12363 times

Back on topic. As for places to generate images I've been kind of fond of https://enhanceai.art/

There's no censorship present there, and they have an abundance of models to choose from. Flux Uncensored is kind of fun to play around with.
That particular model does well with a single character, add a second or complicated poses and it tends to go venturing into nightmare territory.
It's also rather fond of keeping characters nude, so if you want to add clothing you have to be very detailed in your prompt about that (and it will still probably toss some nudity into the image)

I ran the prompt from above just to show some examples. As you can see it totally ditched the comic style and kept it photorealistic which is probably what the model is trained to do.

ff12c87e-769e-4f58-a3c5-f5a565b6d9f8_3.png (1.03 MiB) Viewed 12363 times

e94dff68-95a0-48db-92d3-22f2ed72ba9d_4.png (1.2 MiB) Viewed 12363 times

790f04f3-273f-4141-8a5a-b2615a089320_2.png (1.09 MiB) Viewed 12363 times

[DonShip]

elfa 1.pngelfa 2.pngelfa 3.pngelfa 4.pngelfa 5.pngelfa 6.png
Sorry for the delay. I had some personal issues to deal with. I still believe a DAZ model would be a better fit. The AI lacks sensitivity.

Excellent warrior elf babes. One thing I've notice on almost all, the elf ears are HUGE. Very LONG. I prefer elves with Vulcan ears. Human sized, but pointed.

But their faces and bodies are exceptional. Thank you.

[DonShip]

The only thing I use Krea for currently is to enhance videos, which is especially useful when I'm putting together multiple clips since it does a good job of hiding any color/tone variations between the different clips.

For example. This is two clips joined into one.

ezgif-2-11db665c2e.gif

Back on topic. As for places to generate images I've been kind of fond of https://enhanceai.art/

There's no censorship present there, and they have an abundance of models to choose from. Flux Uncensored is kind of fun to play around with.
That particular model does well with a single character, add a second or complicated poses and it tends to go venturing into nightmare territory.
It's also rather fond of keeping characters nude, so if you want to add clothing you have to be very detailed in your prompt about that (and it will still probably toss some nudity into the image)

I ran the prompt from above just to show some examples. As you can see it totally ditched the comic style and kept it photorealistic which is probably what the model is trained to do.

ff12c87e-769e-4f58-a3c5-f5a565b6d9f8_3.pnge94dff68-95a0-48db-92d3-22f2ed72ba9d_4.png790f04f3-273f-4141-8a5a-b2615a089320_2.png

I don't really have an interest in doing video right now. I'm just playing around creating hot warrior chicks. Maybe someday It'll get good enough to create super heroines from scratch. I am really into costumes, so I see AI's inability to create distinctive costumes a major fail.

Unfortunately, I don't have the time or patience needed to learn how to render. '
Thanks

[VegaTaxeca]

elfa 1.pngelfa 2.pngelfa 3.pngelfa 4.pngelfa 5.pngelfa 6.png
Sorry for the delay. I had some personal issues to deal with. I still believe a DAZ model would be a better fit. The AI lacks sensitivity.

Excellent warrior elf babes. One thing I've notice on almost all, the elf ears are HUGE. Very LONG. I prefer elves with Vulcan ears. Human sized, but pointed.

But their faces and bodies are exceptional. Thank you.

That is so much a +1 from me.