Home Artists Posts Import Register

Content

At the moment I am actively working on Bluethebone esthetic (80s style), or rather, I remove a huge number of watermarks and subtitles from art, I clean up prompts along the way, there are about 200 arts in this lora, I do it right during work and after, almost everything I spend my free time on this, I figured that if I don’t die, it will take about 5 days, then processing and proofreading prompts, and a new model that will practically no longer require negative prompts associated with the text (which still didn’t help) will be posted on patreon с:

By the way, do you have any suggestions? Maybe I should focus on a particular model, or keep developing everything and posting new ones? What would you like to see in the future?

Comments

Snap47

I feel improvements on existing models are always going to have a diminishing return where they would be less perceivable after each iteration. So for all the reasons I can think of, you should broaden your offerings as much as you can as to gather more experience from other models and potentially attract more people with different styles/concepts. I won't be disappointed either way. :)

OneRing

In fact, I'm still learning and almost every day I learn new details about the lora, and I still believe that I make mistakes in training, overtrain the models for greater "accuracy" which affects the quality, but each of my mistakes teaches don’t allow them in the future, I don’t just rivet a million loras according to a template, but I try to understand every time how to do it better and better, maybe someday I will know Zen, buy a better video card and switch to high-quality safetensors models (ᕗ ͡❛ - ͡❛)ᕗ