In the post he shows in the video it says ( ) multiply, and [ ] divide so [ ] is used to make words less important. I don't know about { }.

That would be great! I tried to use the seeds there, but nothing similar to this comes up.

Looks like Daz models... and even though I fucking LOATHE AI... hate it with every fiber of my being... I'm actually okay with it being used like this because all he's doing is iterating upon his original art by upscaling and toning it. There's still the issue with AI being trained on artists who didnt give their consents work, but this is the most benign usage of an AI algorithm I've seen aside from placeholder art. Essentially, this is AI being used as a tool, rather than a vehicle for theft and other scummy shit.

In that case just use repeat. If you specify a number after repeat like "repeat n" then it will repeat "n" times.

Same broo any good recent achieved attempts recently?

try the GPU notebook, sometimes TPUs are not available. if it doesnt work on the first try then restart runtime it should work

No problem and thanks for the feedback, will try the same.

Thanks

Use the stable diffusion 1.5 model, it's the default one.

yep true

yeah. you can use automatic111 instead.

you can use image editing tools to adjust exposure

yep its free, its from a software called balabolka

yepp sure when it's finished will upload

img2img. i used a 3d modelling program to create a 3d char with different poses. then used img2img to get the stylised characters.

restart the runtime in colab and try. if it doesnt work use the latest version of the notebook here: github.com/TheLastBen/fast-stable-diffusion

yep they added paywalls now. you can try using automatic 111 colab notebook.

if it didnt work using mage space for you try it using automatic 111 colab nb...it'll surely work

cool, good to know

ok, try other alternatives like automatic 111 colab notebook. use img2img with the pose you want

one thing i found out thats kinda odd is all the images that people have, if they have the word for the item ur tryingto remove or not show, usuually put that word in the negative box and it should work

use negative prompts or img2img with the desired pose

yepp

There are videos showing you how to train your own model, using another model as the base, and you can make the output consistent. So you could train the AI to handle multiple distinct characters you can target by prompt. Could also handle multiple looks, like street clothes, costumes, etc with multiple prompts. Or you could use a celebrity as the prompt. About half the time the images won't be close enough to the celebrity to be obvious, especially if you pick someone obscure, but the results will be reasonably consistent. Bigger issue I have found is stable diffusion is terrible at props. For example, if you try to do something specific like "swinging an axe" or "holding a bow with an arrow nocked" the AI completely spazzes. And if you want specific angles, or expressions that is also difficult. Can get around some of that using image to image, but honestly I think artists are better off doing those panels themselves for at least a few more years. AI could be useful for backgrounds, shading, and any simple panel that doesn't need a specific view point or expression.

yep, stable diffusion model license is Creative ML OpenRAIL-M and this mage space tool license is CC0 1.0 Universal Public Domain Dedication. feel free to use it commercially

yes you cant copyright ai images

try using prompt templates: github.com/Dalabad/stable-diffusion-prompt-templates