Avatar Studio is now fully open-source, with the fully permissive MIT license. Fork it, contribute to it, make your own commercial product out of it. Whatever you want! I will be actively following submitted issues and pull requests.

Neat. I'll take a look tomorrow for anything that stands out as improvable.

thank you! I hope you don't mind if I fork it to make a full local version

It's MIT licensed, you can do anything you want!

im gonna get a little jiggy with it then

Have fun! If you want to bring any of it back to the original, put in a PR!

I will fork and attempt a PR if I don't go to out of scope I have that problem

Awesome, will try it out later today

Interesting hybrid of Python and JScript. I typically do not DM, but I am sending you one as there may be something unintended in the GH repo. If I am wrong, please feel free to ignore it. I am cloning and reviewing. Thank you for the interesting take and solutioning.

Yeah, the frontend uses Javascript, the backend uses Python, and various tool scripts I made use Python. I prefer Python over JS backends

I generally do too. But JS is better for UI. Lol. Function over style. Don't ask me to do UI design. But with Python, even with the 2.x to 3.x (to current) updates, I don't typically prefer Python for ASync & Threading as they can be messy. Sometimes needed, but the handlers have to be coded and policed just right. But Python does rock for modeling and flexibility. It really does depend on the case. In a pinch, I'll throw C or Go as a light weight web server and do a containerized WS/WGI. But that may be way overkill for the purposes of this. Anyway, props on the release. Good job. Good night. 😎

, I have to admit—this is really solid work. I will apologize if this comes across too much as a code review, but I REALLY enjoyed the repo. There are a few style differences (like adding a quick-start section to the README.md with a link to the details in CONTRIBUTING.md), but that's just preference. The more limited set of options instead of going wide with LocalAI or Ollama makes sense for an initial release to avoid scope creep. Totally fair. The security concepts and code practices are strong. The core ideas are actually very clever and well-formed. It's easy to read and follow, as I'd expect from someone who prefers Python—PEP style basically trains you for that. Lol. It's much better than a lot of the repos I deal with professionally, so I'm looking forward to whatever enhancements you add. That usually takes serious effort, so take this as the highest compliment I can give a stranger on the internet. One quick point, more out of curiosity than critique: I've noticed quite a few tech dependencies that non-technical users might trip over (npm, Python, ffmpeg/ffprobe, etc.). Have you considered containerizing it (Docker/docker-compose) to lock in the requirements and environment? Those things can drift pretty fast.

That's really awesome dude. I'd like to take an evening soon to really test it out, and maybe even make my own experimental fork off if it, see if I can take a crack at it.

Say, if I already have several images of my current avatar that i'm using for sort of dynamic PNG stuff, would that convert well?

It requires the character to be on a solid color background, so you might need to edit it first

Seems easy enough.

Collab with @NeurosamaAI when?

It's time....