Designing a Anime Game with AI! (For FREE) | Waifu Diffusion Guide

Designing a Anime Visual Novel style video game using Waifu Diffusion and Stable Diffusion.
Stable Diffusion
github.com/AUT...
Waifu Diffusion 1.3
huggingface.co...
Waifu Diffusion VAE
huggingface.co...
#stablediffusion #waifudiffusion #aiart

[@RafaelSales55]

you just made me back to video game developement. I used to suck in art but now I got more confident with AI. I will make a bunch of games

[@frogger2920]

I didn't expect you to ever do anything better than your metroidvania amazing video but this one is even better !

[@SW-fh7he]

Love this video, thank you, please keep going

[@AginoEvolutionHD]

I think the hard part is to make the same character with different expressions

[@SW-fh7he]

Yes, what is the solution for this?

[@pjgalbraith]

This is actually pretty easy. You can use Img2Img at low strength for this, or use inpainting. I tried this out and it worked really well.

[@SW-fh7he]

@@pjgalbraith cool, can you do a video tutorial?

[@AginoEvolutionHD]

@@SW-fh7he if can make on video about this since i have no see anyone doing so far

[@loth4015]

@@AginoEvolutionHD Expressions would not be too hard, just inpaint. The harder part is getting different poses, because if you inpaint, say, the arms, with high denoising the clothes might change with it, but if you set the denoising strength too low to keep the clothes the same, then the pose of the arms won't change.

[@williss1192]

It's interesting but how far can you go with this? A simple VN possible with AI? When I look back at the Gameboy or PlayStation visual novels, it seems like it's really possible with today's AI

[@gabrielaginavora199]

WOAHHHH so cool

[@dreamisover9813]

Super interesting video!

[@litz8272]

Can you make a more detailed tutorial teaching how to install this ai?

[@morizanova]

Thanks for inspiration

[@unity3d313]

amazing！

[@chariots8x230]

Would it be possible to also use this to create comic books? If AI were to be used for comic books, it needs to be able to repeat the same characters & backgrounds in different scenes throughout the story. They have to be the same characters & backgrounds, not just similar ones. It needs to be able to represent the characters & backgrounds in different ways, such as from different angles, different lighting conditions, and in different poses. It also needs to be able to combine characters & backgrounds together to create images according to the prompts. I think it would be too difficult to constantly join characters & backgrounds together using Photobashing, as it’s not easy to properly match up the lights & shadows. Plus comics have a lot of panels, and it would be too much to try to Photobash everything. The AI also needs to have characters interacting with each other, and pose them together (e.g., such as two characters from the story sharing a hug). Both characters need to be placed within one of the backgrounds chosen for the story. It also needs to be possible to change the facial expressions of the characters. To sum up, comics repeat the same characters & backgrounds many times throughout the story, and the main feature for comics would be the ability of the AI to repeat these same characters & backgrounds, while being able to follow verbal prompts on what needs to happen in each scene. If these things ever become possible when using AI, then that’s when I think we might be able to effectively use AI for creating comic books/manga. I think your use of AI for designing games is interesting, and it got me thinking that maybe comics might also become possible at some point. Thank you for sharing 😊

[@pjgalbraith]

This is something I've been wanting to try. In the last few weeks two new projects have emerged to tackle exactly what you are talking about, DreamBooth and DreamArtist, both have extensions available.

[@WhatBeDaPointMon]

It absolutely will be able to handle plenty of these things, and already does if you know where to look. Check out InvokeAI, it can do a lot of what you mentioned.

[@frogger2920]

It would be interesting if you could finish one project at least a little.

[@pjgalbraith]

I prefer to torment people with game concepts they'll never get to play 😂

[@frogger2920]

Yes it's sad 😢.

[@loth4015]

Great video. Can you do characters consistently though? I feel like you can not have hundreds of different poses and facial expressions like in a Visual novel by using AI, or atleast I don't know how

[@WalterSit02]

there is methods like textual inversion and dreambooth for that

[@loth4015]

@@WalterSit02 To train your model with, for example, dreambooth, you already need 20 images of your character to begin with though.

[@anonnymous7009]

@@loth4015 You probably have to trial and error a detailed character prompt in different viewing angles and when you think it fits you add it into the ti

[@WhatBeDaPointMon]

InvokeAI. Let's say you wanted the character to flex his right arm but it's down at his side. You can erase that section, draw a crude line representing the arm and forearm, then a circle to resemble the fist. Highlight the drawn line and run the AI, boom. Check out how InvokeAI works and you'll get the picture. I haven't taken the time to use it myself, but that's the gist of how you can use it to change poses and facial expressions. New AI tools will be releasing and fixing these problems in a stupidly short amount of time.

[@tyhall7330]

Can you please do a in depth tutorial on how to install this please. I'm begging you.

[@pjgalbraith]

Search for "Automatic1111 install guide." There are a lot of videos on KZbin already so I didn't think it was worth making another one.

[@user-nc2hs4rp7l]

자세한 튜토리얼있으면 좋을텐데 ㅜ

[@ilusdn6837]

3 years learning to draw to the trashbin

[@WhatBeDaPointMon]

What a waste when you could be learning how to take advantage of these programs instead. You'd already have a huge leg up, but you're willing to throw that away?

[@RocketdogandSeptembr]

The game looks like trash.

[@mushroomcrepes4780]

looks better than 90% of amatuer visual novels