THE LUMINARIUM

Prev Next Random List

Welcome to the Luminarium!

On this website you will mostly find AI-related things such as AI character cards, LLM presets, snippets of chatlogs and more.
Unless you randomly stumbled on this website, I assume you're already familiar with LLM technology and AI terminology. Expect very little in the manner of guides and documentation, this site is just here to host my content.
As for myself, I use SillyTavern and currently Gemini is the model of choice for me.
Further below is just more information about me, if you'd like to go and explore instead, the buttons above will take you to where you need to go.

Contact

You can contact me at illuminaryidiot@proton.me
My mail is open to anything, be it logs, requests or whatever you have in mind!

Current card backlog: 13 (1 in progress)

About/FAQ

Hello! I am Lumen, also known as illuminaryidiot. I like AI chatbots, but the small segment I started writing right after that ballooned and now it's in it's own little block further down. I also like writing long, rambling statements!
As for who I actually am? Well, the image is rather deceptive because it would have you imagine me being an (assumedly) handsome, youthful anime man/boy/thing. Wearing a suit, even!
However, that is just my chatbot persona. A 102-token fiction, created to let me roleplay as a somewhat average person. And, I suppose, the "brand" mascot now.
In reality the person writing these words is a stumpy anthropomorphic raccoon that has to tilt his head up at the monitor despite his chair being raised to the max level. But on all levels except physical, I am human, I promise. No garbage eating here! (Unless you consider pineapple pizza to be garbage?)

For the last few years, AI chatbots have been my main hobby. When I was deepest into it, there were even weeks when I wouldn't touch a single videogame! Luckily, I don't spend the entire day talking to AI anymore. (Or is actually better? I remember having a lot of fun back then... But then again, nowadays I am mostly into it for the creation aspect, not chatting. There are only so many interesting scenarios you can go through, you know?)
I also enjoy programming, and this site has been my hobby project for the last year. Thanks to chatbots, I finally got to learn a little bit of HTML and JS! And before you ask, I code videogames normally (as a hobby) (C#, currently Godot but will probably move back to MonoGame for the next project). And no, I don't have anything released, because I as many others scrap their projects immediately once things get difficult. Though curiously, it's almost always when it's time to make actual art assets that things start slowing down...

Other than that there is not much else to say, really. I just like making stuff!

On AI

The other avenues of AI content don't interest me much though, for one reason: they are not interactive, unlike chatbots. There are only so many times I can be entertained by someone prompting an AI to draw a silly image or have a voice AI say something a celebrity normally wouldn't say.
While chatbots follow the same structure of prompt-response, the way you utilize them in a continual chat is much more fun to me than the static responses you get from the other fields of AI. The idea that you are actually participating in an active RP with a character/RPG card/whatever is fun.
Imagination and reading also play a huge role in why I enjoy LLMs but not the others. Being able to read and imagine scenes and dialogue is stimulating, and this stimulation is something the outputs of other types of AI generation sorely lack. An AI-generated image of a character waking into the room and an AI-written paragraph describing the same action might be similar, but reading it forces you to imagine how things look. What does a playful saunter look like? A wicked smirk? Or a furry fox-woman frantically juggling 5 eggs all at once trying not to drop them?

As a disclaimer, I do use AI image generation tools for many of my character cards. Why? Because finding images that fit very specific OCs isn't easy, and I am terrible at drawing. To me it is just a tool, which leads me to...

AI, it's use and future (Written 31/03/2025)

It would be a lie to say that generative AI has not advanced massively in the last 5 years. But where will it go? Will all human jobs stop existing in 10 years? Can artists thrive in a world full of AI-generated media?
In short, I am cautiously optimistic. As with any new tool, there will be people who say the technology will completely take over and endlessly improve (the concept of flying cars comes to mind). I think AI, in this current iteration is limited. And that we are starting to brush up against their limits.
Many of the major LLM companies these days are completely focused on creating assistant AIs, giving up on the idea of creative writing or other uses. I also see the introduction of reasoning models as proof of stagnation, the fact that they would slap a CoT (forced, mind you. Can't trust the consumer to decide when and what they want the model to think about, do we?) on their models and act like it's something new and revolutionary shows to me that they are either lazy, or that they do not have anything better to show as innovation. This is not to say CoT prompts are bad of course, they do work, but my main contention is that nobody built these LLMs with CoTs in mind. It was just something that was discovered.
Right now the majority of LLMs are general-use models. They are huge, slow and cost a lot of money to use, and have been steadily growing bigger and more expensive each year. Not ideal. I think that in the next few years, we're going to see more specialized AI models. They won't be much more intelligent or have bigger databases, but they will be more focused depending on what task you want to accomplish with it.

As for image generation, I think we're reaching a ceiling on that too. AI-generated images are really high-quality these days, and I think we're going to see diminished returns on any subsequent models. In the quality department, the biggest hindrance is that you need high-quality art of the things you want to depict to train the AI with in order to get good outputs. Other than that, I think the next step will be for image generation services to adopt controlnets and give users inpainting/img2img options, which some providers already do.
Artists will definitively exist in the future, as people who were willing to support them through comissions/patreon are not going to evaporate overnight due to an AI LoRA of their artstyle existing. The people who argue that artists will die lack a fundamental understanding of the artist-consumer relationship. People who want more art from an artist will support them if they can, meanwhile people who were never going to give them any money still aren't going to.
Of course, this is with the current generation of people in mind. When people are brought up having AI at their fingertips all the time and are able to just request any art in any style, then the question of "Why should I pay you when I can just have AI generate it for me?" will no longer be a callous insult, but a statement from a new culture that knows nothing of the joy found in creative works. Which is a scary thought. And a situation that I hope will not happen. Do they still give kids pens and paper in school these days?

Video is still a work in progress, and by what I've seen it can still only do floaty, slow movements. Any rapid motion will just look like a character is jerking their limbs around. I am not impressed by what I've seen so far, but maybe it'll change in the future. AI-generated movies are a possibility, but I do not know if audiences will want that outside of the spectacle of watching the "first fully AI-generated movie".
To me, movies (and shows) are hugely about performance. The engaging thing is to see actors... act. If you were to make an AI do all that, it would just not be interesting to me. And considering how even mainstream audiences are getting tired of the CGIfests that come out these days, I'd say a lot of others would hold the same opinion as me.

Now voice AI is something I do not know that much about. Seems to me to just be a continuation of voice synth technology. It'll be easier to fake other people's voices. Otherwise, not that interesting.

But speaking of AI voices, let's move on to the final topic on AI, AI video games! Yes, why bother making a video game when you can make the AI write the code for you, and the art, music and voice acting too? It's so easy!
In short, no.
Both AI images and writing need humans to adjust it in order for it to be presentable, I assume music, voice acting and 3D models will be the same. It will still require human effort to make something good and enjoyable.
I do understand the idea that indie developers might not have the money to hire certain artists, but I also believe that people who buy indie games are highly aware of what they are buying. This is my main contention with the idea that AI games will take over the market. For example, you can easily churn out a game with RPGMaker default assets in a couple of days, but will anyone buy it? Probably not. The indie consumers are basically trained to ignore RPGMaker games, unless you're someone who enjoys the few good ones.
In order for any game to be successful, you either need to have a good game, or some good marketing behind you. Now, AAA devs will definitively be able to get away with having AI content in their games due to massive marketing budgets, but who even plays AAA games these days? The people who just want the latest thing and play for five hours a week calling themselves a gamer. Those people are not in the market for good games and do not care. Which is why I'm mostly talking about indies here.
But to have a good game, you either have to have a great team, or be incredibly anal about quality. Or both. AI will never be a great team for you, I'm sorry if you even thought that. But since most of the people who think AI will be able to make games are just ideas guys, they have no idea what quality control even is!
Basically, what I'm saying is that because of how the indie space automatically filters out garbage and clones of more popular games, you will still need to put in a lot of effort to make a successful AI game. That most potential buyers eyes' will glaze over when they hear your game is made with AI (just like what happens with RPGMaker games) won't help either. But good games do eventually find their audience, even if it's niche. I am not saying it will be impossible for AI games to be successful (some FOTM AI garbage game will most likely be made within a year), just that it will not be as easy as others claim. Both in the amount of effort you will have to put in, and the amount of sales/recognition you will get.

Phew. Now that that's over with, let's move on to something more lighthearted...

Currently, I am...

The password is... HOLY FOX