AI [AI Art] - Show Us Your AI Skill *NO TEENS*

[Disharmonia]

 

[Sgt028]

By default Stable Diffusion (assuming it's stable diffusion you're using, not midjourney or DALL-E) generates different characters depending on the pose or seed. In each case you'd need a strategy to keep the characters consistent. Now, if you're using a website to generate your images, I think you'll have more limitations/constraints as I don't imagine they'll let you use controlnet.

If you're using a website, a hint I could give you is to write down your hyperparameters such as positive prompts, negative prompts, model used to generate the images, sampler type, sampling steps, CFG configuration value, width, height and seed.

So, the next time you try to generate new poses, you just change the prompt pose description. Then, use face in-painting in-case you are not satisfied with the similarity and use your image as input.


Now, if you're using Stable Diffusion, you can load your configs by simply clicking on PNG info tab and load your generated image where you have a reference image or image zero and then click on send to text2image, then use openpose ((search civit ai)) and use reActor to automatically do the face in-paint, but do notice it will never be perfectly the same


Watch this video:

You must be registered to see the links

As the wolverine face changes, that is not a feature, but how stable diffusion works, you can search more videos for yourself, they use the extension animediff, you need camera angle tricks to keep the face visibly the same

Thank you, my friend!!! this was of great help!

 

[BeriJuns]

Nude on shopping street

 

[Krysis_here]

 

[Mipecio]

 

[BlenderGuy]

Having a consistent char, that's definitely difficult, I've tried several video tutorials, but still can't get the "perfection" I want, in the end the more consistent results I've got were from txt2img followed by tons of img2img, inpaint and good old Photoshop.

Btw, I would give a look at some foundamental extensions for Stable Diffusion, ADetailer and Reactor, they may help tuning the small bits and faceswapping.

One of the online products I use seems to be consistent with the character produced.
But then to change their body positioning/framing you may need to tinker with adding an additional Lora to the mix. I'm installing stable diffusion today,perhaps that will hold the key.

 

[Bossyboots]

Nude on shopping street

View attachment 3545716 View attachment 3545717

She is most likely shopping for clothes... Or perhaps sunglasses...

 

[BeriJuns]

A famous actress during the shooting of the royal porn film

 

[Mipecio]

 

[Bossyboots]

View attachment 3546113

She is no Angel... XXX

 

[BeriJuns]

Who doesn't love nude catwalking?

 

[Messalina's Pizzaboy]

As promised: Unpublished BTS material from the "Sporty Illustrations Swimwear" calendar session
Part I
Advanced pose striking with Linda Kissmegoodson



Holy s***!



Twyla "The Twins" Oladejo. They were all jealous...

 

[Lemmywinks]

Surfer Girl

 

[BeriJuns]

Experienced nudist enjoy vacations

 

[Leongen43]

 

[BlenderGuy]

Prompts are

1 indian desi girl, (nsfw), (lying on bed, legs apart), long hair, large cleavage, (1 perfect little _vagina:1.4), (huge breasts: 3.1), (black stockings:1.2), beautiful face,

(8k, RAW photo, best quality, masterpiece:1.2), (realistic, photo-realistic:1.4), (extremely detailed CG unity 8k wallpaper),

Prompts are kinda messy tbh but I'm doing so many other things today.

 

[Krysis_here]

⛆ Wet and Beautiful ⛆ ​

​

 

[Leongen43]

 

[Mipecio]

 

[Krysis_here]

☘︎ Colorful Hairs ☘︎ ​