End of 2023

Hi, all! Yes, I’m still alive, and still grateful for those of you who are still supporting me in my endeavors, as meandering as they may be (hyper-distracted rabbity-thing and all that…)

A part of me doesn’t want to check to confirm (don’t think I want to realize the truth), but I want to say I’m about a year into this hiatus from Interns. It’s really invigorated a lot of my creative impulses, and staved off complete creative burnout—something I’ve experienced before, and desperately wanted to avoid.

Lately, I’ve been feeling the itch to start tackling a project of my own to some degree, but I’m a bit conflicted over what that should be, or where to start; a part of me is wary of committing to anything that will give me that “trapped” feeling, so I want to take things a bit slow as I do that.

I’ll save my thoughts on that for a bit lower in the post.

For now, I’ll talk a bit about what I’ve been up to, and what I’m still planning on working more with.

AI Stuff
I’m still very much working my way down the AI rabbit hole (ha!). I’m well aware that a lot of people have really strong feelings in regard to AI stuff, but, I see it as an exciting new series of tools that may (hopefully) enable some less-talented and motivated people, such as myself, to be able to do projects that would ordinarily be beyond our abilities.

It can definitely be abused—and I’ve been seeing plenty of examples where it’s been employed to throw out some quick artwork with little time being invested to correct blatant errors when it’s obvious that the piece isn’t intended to be a placeholder of WIP image.

There’s also a tendency for everything to look the same, as people tend to find one style that looks good, and everyone else copies it.

Personally, I’m hoping that I’ll be able to find a way to combine AI with HS in a way that will let me create something a bit better than with HS alone.

AI “Filtered” 3D Models
One of the big issues with using limited 3D software is that I’m restricted to whatever elements others have created or imported models for. There have been more than a few times when I’ve resorted to using my image-editing skills and Photoshop filters to attempt to create things that I couldn’t get in 3D. There have also been some cases where the HS model anatomy is so abysmal in some poses, that I’ve either needed to just let it go, or spend hours trying to correct things as best I can.

I’ll be sharing some quick samples of some recent experiments with using AI like an image filter, to attempt to get a little more life (and better anatomy) out of the somewhat lifeless HS models.

Above, you can see the quick attempt with ‘May in the arcade,’ and there’s also this equally-as-quick take on Chastity:

With Image-to-Image generations, you can specify how much power the AI is given to change the initial image (with a range of 0.0 to 1.0). Too little, and you just get the same image. Too much, and it completely ignores the original.

Things can quickly get out of control…as evidenced here:

There are tools to help with this, however, like ControlNet, which has a number of different ways to influence the AI generation.

In the case of the May image, here are a few of the ControlNet models that I played with while trying out the “filter”:

And here’s a plot of images created with the AI denoising strength going from 03. to 0.7:

Image editing is still helpful, of course, as I can mix and match parts of different generations when trying to build a final image.
Obviously, more work is required if I want to make sure I actually get things like reasonable hands and such…as AI is notoriously bad with those.

LoRA Training
So, over the last few months, I’ve put more effort into AI training, with more of a focus on LoRA training, as I’ve been getting a bit better at it. With AI training, I should *hopefully* be able to get the AI to understand concepts like what a character should look like, or define an art style that will help my creations stand out a bit.

I’ve been working off-and-on at creating a style LoRA for X-Change Life, attempting to match that game’s mix of AI filters, image edits, and Midjourney creations. While my latest attempt at one manages to match the style more closely than any before it, I really feel it acts a bit like a sledgehammer—turning *ANY* image into something similar to what exists in the game…to the extent where it ignores much of the original specifications of the prompt.
The image below shows it being used at increasing strengths, with the first image being what the AI was asked to generate, and the last being the LoRA at 100% potency.

Ideally, what I’d want is the first image, except with the style of the last, but I’m clearly getting more than just the image style.

I want to save at least a little something for the backer post, so I’ll include some more samples of images made with this style LoRA there, along with some NSFW samples from a different art style LoRA I’ve been working on (that failed miserably).

Modding

One of my other big focuses since last time has been modding for XCL. I was *hoping* to expand the bar-drink-spike mod I was working on with a little bit of story and gameplay mechanics, but it ended up getting completely out of control and turning into a bigger project than I had originally anticipated.

I’m still hoping to get it into a playable state and pushed out the door for others to enjoy, but I keep finding errors and aspects that need to be fleshed out. So far, though, working on it has been the closest I’ve come to doing actual game development in a while—both the good and bad aspects of it (as I’ve come to resent the mod a few times while working on it).

Still, getting back to my roots with game modding has really reinforced the point that I probably need to work on games that I can play and enjoy myself, and I’ve seen I really like seeing the things I’ve created pop up during gameplay. So, this is clearly something I need to keep in mind when deciding what I want to do next.

Moving into 2024

So, that’s some of what I’ve been playing with—and, of course, I’ll share some of my additional AI generations in the backer post, though it’s not as much as previous. Here’s what I’m planning on doing, going into the new year:

My AI To-Do List

Art

There’s still more I’d like to spend time looking into with regards to Stable Diffusion. As it is, I’m behind the curve, as I’m still using Version 1.5, while I’ve seen other people getting much better anatomy out of SDXL.

There are a few additional tools I’d like to spend some time exploring, such as ADetailer, which can identify elements of images (faces, breasts, genitals) and make corrections automatically.

I also remember coming across a plugin that allowed for drawing a segmentation map—dividing up the image into different areas that could take different prompts. This means that I could describe elements of the image in sections that would keep prompt elements from “polluting” different parts of the image. Likely handy for describing two different characters in the same scene, or background elements.

I’ve been seeing a lot of people having great luck with using face-swapping AI to help keep characters on-model, too. Seems like it could be useful to have a master character document that could be used to help fix images that are veering too far from what they’re supposed to look like (there’s also a “Reference” ControlNet model, too).

Writing & Code

There’s the LLM/GPT side of AI that I still want to work with, as well. During my hiatus from Interns, I’ve had plenty of story ideas, but what I’ve been struggling with is actually sitting down and writing anything; I’m still not sure what is causing this particular block for me, or the best way to get past it.

I’ve had plenty of suggestions on how to use GPT as a writing aide, and I’ve seen some potential from some experiments (such as using my story samples to help create style guides for the AI to work with, whether novel-style or how I’d write a more traditional ADV-type visual novel).

I’m a bit curious to try something like a connect-the-dots approach, where I’ll write story content, put in some notes whenever I start drawing blanks or getting fatigued, and have the AI try filling in the in-between sections of the story that are necessary, but not as exciting to write.

I’d also like to go back and give Auto-GPT another try, now that I have a second system that I could run it on. I’m still curious to see how well it would work as a coding assistant, if I was able to give it a design document for it to work from, with the self-running aspect giving it the power to do multiple passes and expand on its own work.

U666 Game Project

As I said at the start, there’s a part of me that’s itching to create something all my own again, but, admittedly, I’m torn. I have plenty of ideas, but deciding on what would be best in the long run is challenging.

Interns - Well, let’s start with the obvious one. Picking up Interns again is my first option, and I know that’s probably what most people will want me to do. I’ve certainly been thinking about it a lot, and I feel that, if I do restart it, I’ll likely take a game-mechanic-first approach.

As I’ve been reiterating, I think working on a game will be most rewarding for me if I can actually enjoy playing the game myself. Interns was always intended to have a semi-life-sim aspect to it with sex, willpower, and mind-control elements, so, armed with GPT coding assistance, perhaps I can tackle the core gameplay loop first, and revisit working the existing story sections back in, once I’ve got something I think is enjoyable to play.

“Short” Story-Based Game - When I first started playing with Stable Diffusion, I thought it’d be an interesting idea to use it to create artwork for a quick HTML-style branching-story-style game. AI art being new and novel, it’d be interesting to see how it would be received.

Of course, it’s been months since I actually *had* that idea and games with AI assets are everywhere now. Guess I was too slow. But the basic concept I had for the game still had a decent amount of fleshing out, so I could always do *something* with it.

Combat-Defeat RPG - CoC was my gateway from Oblivion modding to text games, and I’ve always wanted to make a game that focused more on the elements I enjoyed from the game. So, something less furry and waifu-centric, and more on exploration and combat defeat.

I have a sci-fi-based concept in mind that I’ve been fleshing out for years, and I haven’t gotten completely bored with it yet, so it seems like it could be one that could hold my interest for a while. It would, unfortunately, be very challenging code-wise.

Standard Life Sim…but Sci-Fi! - Another game concept I’ve been playing around with for a long time that actually came into being after I realized that a world I had created for an Alien-influenced story would make for a potentially interesting change of pace from the usual life sim game.

Rather than being someone who has just started exploring a new city, earning money, and meeting people, you’d be doing that…in space! …on a new colony world. The typical life sim mechanics would be there, but also space-drugs, body transformations, and the like.

Anime-Style Visual Novel? - This goes way back, but I’d originally been inspired to make Interns after seeing games being made by repurposing assets from Japanese visual novels. Those had the same drawbacks as real-porn games, however, as you couldn’t make anything that you didn’t have media for.

I had tried using AI over top of anime-style models before, but I’ve learned a lot since then, so it might be more doable now. A simple 2D art style could be easier for me to make changes and corrections to.

I do have some partial stories that I’ve worked on, taking place in my “Change Crystals” universe.

Anyway, I’m just tossing out a few of the ideas that have been rattling around in my brain for the last few months. I keep telling myself that I need to put together a “story concept sampler” project one of these days, that would give people story samples from each of the ideas to see which ones folks would be interested in, but I always get hung up on the presentation (I’d rather not just dump out text files—that’d be boring.)

Anyway, thanks for reading, and I hope everyone has a Happy New Year!