--13/11/2025-- update

been making some progress on my 'ai-cc' project but even more on the research side of things , you see I finally gone into the LLMs rabbit hole, and lemmi tell u Its quite a fun one to get into I've run a couple of RP models , learned how to write a somewhat decent character card is was amazing .. BUT .. I ended up falling into the scope creep trap ... again 🙄

you see as collector at heart (the kind who spend hour organizing their minecraft chests just for fun) I wanted to see my ai things neatly index and organized sooo .. i might have added this

its an ai models/resources/things manager i dont what it'll end up being called , but think of like the lora extra networks tab for LLMs and their stuff, mby even TTS resources if ended up collecting them it should act as a sort of local hub for consulting ai stuff collection as well as run them though backends like Ollama or KoboldCPP

also started a an RP character builder page that's can be exported into SillyTavern character format

these was also a day where i played with color scheme a bit and refined the old theme as well as made new ones like this Cyberpunk one

--30/10/2025--

-- some of you might've noticed that been somewhat absent these recent couple of days, that's because I've been putting all my free time into a pet project lets call it "ai-cc"--

About:

its basically a character creation tool using SDXL (Illustrious specifically), I remember back in October 2024 releasing this wildcard collection for PonyXL thinking to myself wow today's model are so accurate that they can actually be a full on character creators! and making this pic to go along with it

now guess what its October 2025 full year since then and models like ILXL have gotten even more amazing so im taking the dive into making a full on character creator this time!

The concept:

the concept of this creator is the basically a prompt formulator with a bit of of twist to it, while formulating the final prompt it keeps tabs on three variables [Trait stack, Outfit stack, Prompt space]

Trait stack : is the all the physical traits that character have stuff like body form , eye shape and color ...
Outfit stack : is assembly of all the worn items that character have from head to toe
Prompt space : is basically a visualization of what prompt should cover basic on the current view conditions like position, rotation, zoom.
- say for example the character is facing backward while the Trait stack still have properties relating to the character's eyes , the tool would filter those as they're not logically visible within the current requested Prompt space
- another example is when doing a face close-up while the Outfit stack have stuff like footwear set, the tool would still keep tabs on those and hiding and showing on need based on the currently set Prompt space when it formulates the final prompt

Planned features:

core stuff
- a user friendly game-like web based interface
- image generation with metadata in them
- and extended database of most of character options promptable with illustXL
- exporting/importing creations from a file
further expansion horizons
- a character library
- a storyteller module (creating scenes with changeable characters)
- expanding on the character for using it with an RP LLM
  - a character profile builder for RP like (background, personality, skills, stats...)
  - a simple llm chat interface (im thinking something VN like)

--Note: I know close to nothing about ai RP and local LLMs so i you have any info that could spear me hours of chasing leads about the subject I'll be rly thankful--

Choice of tech:

at first I'd thought of building into Webui like i did with my Wildcard Gallery Extension then i quickly scrapped the idea, I dont like the thought of additional tools being load in when not needed and I don't like spending time delving into A1111 code to see how i can get around things also I like to use something other then gradio3 for a change.

thats why I'm making this tool standalone communicating with A1111 via local API this give me a lot of freedom to code around and peace of mind about compatibility when switching to another A1111 fork

I'll be building it using NiceGUI for python which I've been wanting to try out of a while now

Using SDXL for assets:

Luckly I stumbled on this amazing lora for ILXL it iterate quickly on clean monochrome assets [some posts here].

I generated alot out output using Txt2img , roughly spliced them together in photoshop and then put them back in img2img at 0.7 denoise with the same LoRA and a another prompt to get somewhat consistent assets something like this