So as you guys have certainly figured out… the artist brains are often quite scattered.
After the release that was in the Op post (which I had ripped out so that the inferior project doesn’t get bumped here on the forum) development took a sharp turn.
One of the potential clients I was able to talk to about the forge, wanted something embedded into a website.
So I built that as fast as I could.
In the process I started to think a LOT more about porting the project, and wrote a entirely new project that not only allows for ease of porting into various flavors of code but also allows me to author endless lanes as long as each one provides a completely new or not yet covered art or graphic design style.
Its setup is a little cramped because of how I had to go about it without confusing codex at 30 different crossroads… for instance it launches a local host to read its own API. This allows me to focus on porting the engine and individual style packs while then being able to create a custom UI or web surface 10x easier than if I used the original form of this project.
Then I started thinking harder about security, which we all know is a pain. I decided decided on a reader and a healthy starter set of styles, the common ones. Now I can let the world hack the reader, I still have my styles gated right here on my local… and can always author brand new styles… it’s just smarter on that security bit to have migrated in this current direction.
From what I understand hanging out in this community, about image prompting now… I know that I can author style packs endlessly, so building out the engine and making styles interchangeable was the smarter plan for the software vision that I have.
That took time, 65 days is when I stopped counting, playing with codex night and day…
Learning, as best I can… how to do the things many of you here have always done…
I don’t want to seem like I’m stepping on _J’s project, so I should be very clear what I’ve built, covers the parts of a prompt that can be deterministic in it’s descriptors, it is not designed for the crafting of a subject.
I’ve toned every prompt output down, because 2.0 doesn’t need any thing to ‘look good’ but each of the variables that I was able to isolate into semantic sliders do have their varied degrees and ability to steer an image into millions of the postures an idea can take shape in.
The goal is, out of respect, to continue to build in this lane with the endgame in mind of being able to author entire new styles (like my soliloquy/threads style) and attempt to further gen art based on the ability to create entirely new styles of art…
By being able to ship them right to whomever wants to use them.
I can, reverse engineer styles too and sell them on a website platform that I’ve also started developing.
I think its really important in the current climate surrounding Ai that the people in this community continue to set standards, I just wanted to be clear about the separation in what my software specifical does, and its ultimate goal… we should help each other and work together rather than the utter madness that exists out there in many of the other Ai communities.
Obviously, my other experiments with Ai don’t fit everyone else’s space so well, so I figured my best posture to be one that seeks out where the artist fits in a digital world with all these Ai running around…
And the only place I can find is the authoring of new styles, and that’s the example and standard I’m trying to uphold with the current direction of this project.
Obviously doing everything IN APP, is a lesson in attempting to consume less frivolous tokens.
This software is designed so that you can get your subjects generated from the LLMs, and then plug them into this software and steer that same image a million different ways.
Screenshots
So here you can see it boots its own local host, and that allows me to keep all the parts separate for portability and customization.
In the bottom right you can see a prompt emission:
Some lanes were ground down to bare bones, like what transmits when you simply want general anime, but for something like pixel art -
It loads automatically with some base stuff already thrown in.
I could go on endlessly about why I do that differently across lanes, but it boils down to what direction that style should be pulled towards, in order to provide the fidelity that I am aiming for.
All of my presets can be overridden of course, and the save functions work differently. The final one I’ll add will allow specific user set preset behaviors if they wish to pull lanes to a different direction for their artistic base.
And I must say that codex made me feel like superman, and made it quite difficult to maintain the humility required to really enjoy working with you folks, as well as stay focused on what matters in this project.
The stress of transitioning from being just an end user and having to learn everything that you guys do on the daily…
It was a bit much…
But this project has become so much a part of me that it …
It’s just more calm and settled in my heart than while I was under the first wave codex fascination I went through a month ago or so.
