mrsilverlust
patreon
More thoughts on AI and answering your questions
🕑 Added 2022-09-18 12:38:00 +0000 UTCAlright, here is another post about AI that no one asked for :p
“Where may we find the AI?”
I think this is a good place for information:
https://www.reddit.com/r/StableDiffusion/wiki/index/
To run it, I like this fork:
https://github.com/AUTOMATIC1111/stable-diffusion-webui
If you have a good gpu and can follow the instructions under “Automatic Installation on Windows”, you can install and run it locally. Otherwise, if you have a google account, you can try the google collab from the same link. Just run all the commands from top to bottom and open the generated link in a new tab.
If you just want to try it very quickly and don’t want to bother with all the complications and settings, try here:
This is just made available by a random guy that did not expect the amount of users, so he might have to take it down soon. See if it’s still online when you read this :p
Sites like this are particularly cool because you can quickly use them on your phone.
For tips, and other info:
https://www.reddit.com/r/StableDiffusion/
To browse other people's creations:
“Do you think it would be possible to use such graphic in VN games?”
This was already answered in the comments. Indeed, we aren’t even close to that yet simply because the AI cannot create the same characters twice in different scenes (unless we are using as a prompt a public figure face present in the training data for the AI). Besides that, even if you create some beautiful pictures:
You’ll see that the AI still makes too many small mistakes: the girl's left eye in the first picture, the bones around the neck and the clothes in the second, and the eyes and face in the 3rd also look a bit out of proportion. (note: this was supposed to be the same character, a blonde girl with blue eyes, with the same style..., but you can see how different the results are)
I don’t know how long it’ll take for the technology to overcome these limitations, but it’ll get there sooner or later.
“Regardless, it is time to not merely put away so childish a thing, but to smash and burn your new toy lest it steal away your soul.”
Some people, rightly so, are afraid of the effects of what the AI can have on the world. Not particularly this one, but what it’s to come. One way or another, it’ll change everything, probably even to a larger degree than what the Internet (and then smartphones) did in the last decades.
But, right now, it’s just a tool like any others that can be used by anyone to elevate what they can do. Really interesting video here showing how an artist can still use the AI as a tool to get better results than a “normal person”:
https://www.youtube.com/watch?v=NiJeB2NJy1A
And here is someone explaining how the AI can be used, at this very moment, to help with designing icons for their game:
https://www.youtube.com/watch?v=blXnuyVgA_Y
This is quite interesting. For example, in the 1st chapter of Nothing is Forever, you’re asked to choose a logo for the Neural Fitness company. While one of them was made by an artist under the creative commons license (1st in the image bellow), the other choice was made by me that have no skills whatsoever to design a logo (2nd in the image bellow).
But now, I can ask the AI to generate me some nice logos of a brain exercising. The results (all the other images) aren’t better than what the human artist did, but can be better than what I did (with the caveat that the AI cannot write a single thing to save its life, so, forget about logos with letters).
This could also be used to give a bit of more individual styles to the individual characters:
Those logos look like the kind of thing Lea would wear. Perhaps I can use it in one of her outfits?
Unfortunately, this particular outfit I was playing with has the wrinkles baked in the textures, which makes it hard to overlay the rabbit image in a realistic way and I didn't want to lose too much time on this test (it doesn’t help that I know nothing about fashion xD). But you get the idea. Someone a bit more talent than me (or with more time) could come with very cool results.
Another use could be to help populate scenes. It’s quite hard for me, due to constrains on hardware and characters, to populate a scene to make it look more real. Could AI help with that?
For example, this render will be in the next update:
Could we add some background characters?
The answer was no. At least not in a quick way. It took me too long (it would have been faster to do it the right way on Daz) to get just that one person to look minimally good, and the perspective and proportions don’t even look right.
It may be just me being a noob with the AI, but I don’t think I’ll be using it like this any time soon.
But perhaps we can still use it to bring some more character to the environments? Maybe to create some original textures?
Just due to the topic, I was thinking that a friendly robot graffiti would be cool:
What if we add it somewhere to the city?
In this particular scene it doesn’t benefit immersion that much, but we can think of situations where it would help a lot to bring an environment to life. Even just to customize a character's home with the kind of paintings they would like and things like that.
Conclusion: The future possibilities are exciting and only limited by our imagination, but I won’t be using this AI any time soon. Either due to limitations of the technology or due to my knowledge on how to consitently create good results fast, I don’t think it helps me much right now in my work-flow. But it’s definitely something to keep an eye out for.
And with this, I promise I won’t talk any more about AI and the future posts will be about the game :p
Cheers
