- Apr 14, 2023
- 453
- 443
- Apr 14, 2023
- 453
- 443
bro u sick as fuck how disgusting
have to make friends with you later (secretly)
btw is that platform free can u send link i have lots of idea
I'm using Stable diffusion with my GPU. If you don't have a good GPU, then some free alternatives include going to Civit.ai or Yodayo and using their image generation services. For reference, I used the model Hassaku Hentai XL, Better Eyes Version to generate the initial image, then did multiple img to img steps to clean up and refine certain parts of the image. Have fun!
- May 19, 2020
- 7,619
- 14,637
From the way he wrote it, it looked like he was trying to spread a bunch of that anti-AI bullshit.
You gave me another chance? What do you mean by that mister?
That is a bold faced lie and you know it.
This is beautiful! I am so happy you took the time to make this and so quickly on top of that. Ngl, a ToT remake with artwork of this quality would be amazing. I am going to make this creation my next wallpaper!
I don't know anything about making AI art but when my friend visits again I am going to show them these prompts and go from there.
Yeah, experience with using the AI is making me get quicker at reprompting for what I want and fixing imperfections in the image.This is beautiful! I am so happy you took the time to make this and so quickly on top of that. Ngl, a ToT remake with artwork of this quality would be amazing. I am going to make this creation my next wallpaper!
I don't know anything about making AI art but when my friend visits again I am going to show them these prompts and go from there.
The image you saw was not sourced from any Toy artwork, lol. It would be cool if the game actually had a scene to denote the start of the shoe game - that's just what I imagine happening.
Also keep in mind the legality of making profit from artwork purely done by Ai is still in question, even if you do spend hours of editing and resources to get a result that isn't filled with Ai nonsense. It's a tool, an extremely effective one that reduces the training and skill required to create works of quality, but a tool that requires effort nonetheless.
Until the AI craze goes down and people start seeing Ai for the imperfections it has and stop attributing all the success for the images you see to the ai, rather than the person working on making it good, then I don't see something like that being viable for a paid game.
We've gone from people asking the most obvious questions ever to literally
"Guys how do I cum IRL while playing?"
Jesus christ.
"Guys how do I cum IRL while playing?"
Jesus christ.
Is it possible to access the eighth floor in the game version 1.18.4? And if it is, how do I do that?
nope the scenes are mostly attacks or out comes of certain battles
tbh ToT became more than a fap game ppl alrdy told me that when i first started i tought it was bs but it isnt even if u view it as fap material u will get attached to the story and the characters and for scenes i rlly think they should add a gallery but i also think it make the player less intressed in grinding
- Mar 9, 2022
- 17
- 12
This is amazing. Would be great if the ai could have a prompt that applied to everything in the gallery so that you could mass convert them. I always thought you could make about 10 with specific prompts and give the ai those 10 then also give it the 700 or so other images from the gallery and tell it to use those 10 as a basis for what the remaining 700 images should look like.
https://f95zone.to/threads/modifications-fvlaenix-modifications.117902/
on this page are links to patches for different version of TOT and other games and within them are gallery patches that allow you to access all the scenes from each one of them.
I mean technically you could group images into sections and use that logic to apply image to image work, butttt... Prompts are pretty important.
At low denoising levels the image won't change much and probably get worse, at medium denoise levels a prompt not fully matching the denoising patterns of the image will often generate eldritch ai nonsense, and denoising at too high of a level without specifying exactly what you want will get an image so far removed from the scene that you likely wont recognize it (along with a little ai nonsense).
You technically could do such a thing, but you'd be using your computer resources wastefully for a result that isn't even good. For fap quality results (minor details will be off and some eldritch nonsense in every other image) you could prompt like a subset for every scene, and then go without editing the image. But honestly if you're just going for fap quality, you should skip img to img altogether and go with a raunchy prompt.
Ideally you would gather a bunch of images and use them as a dataset for training loras, which can be a real pain. In regards to prompting, due to the way tokenizers work, less is quite often more with AI's. Oversaturating a prompt is not only detrimental to artistic design, but it takes away from loras and modules like bad hands, easy negative, etc. which results in "common AI issues" like extra/missing fingers, or a head growing out of somebody's shoulder.
It is definitely possible. What you would need to do is build a workflow with comfyui. Shouldn't even be that difficult.
- May 25, 2022
- 31
- 86
I wasn't aware that lengthening the prompts causes more "common ai issues," but a lot of times in regards to specific characters like this game, it does matter. You have to spend a lot of tokens to properly prompt things like the outfit, character traits, action, background, and so on. Without those things persisting through an img-to-img process, you'd essentially "decharacterize" your image.
Already some level of decharacterization is inevitable through detailed prompting, different models have different styles and you lose the original style when you make a significant enough change to make the image look better. As I said before this, 'technically' yes you could do something like that, but... I'd say according to my experience with AI the results likely wouldn't be true to the character, even if low levels of prompting avoids AI nonsense.
I'll try it a little since I'm curious, maybe what I'm expecting isn't true, but don't get any hopes up.
Last edited:
As for Loras, I'm not familiar with them as I've never trained one before. From using them however, you do get a 'character,' but it comes along with the art quality and style of the images the Lora was trained on. All we've got is the TOT images to work with if we want to create a Lora for say Scarlett or Kaori, and low quality input is just gonna give low quality results.
Try homogenization and build from there. Batch convert images to black and white sketches, batch edit image to be more anime/life like, batch convert images to anime/cartoon. Repack all images into rpgmv files, and bobs your uncle you've recreated TOT with AI high quality assets.
Edit: The purpose of a lora is not just for character. It's to train the model what it should do with certain key words.