This Lora is intended to improve my experimental AnicomiX Checkpoints:
ALPHA / BETA / GAMMA
I'm reaching out for your invaluable support! To accelerate the improvement of this model, Feel free to experiment with various backgrounds, ethnicities, hand poses, facial expressions... I would be incredibly grateful if you could share as many of your creations as possible right here. Your contributions will be a goldmine for my future training sessions and will play a crucial role in refining the model (and this LoRa).
Graphic Style between Anime and Comics
This checkpoint is an exploration in creating and improving AI models for image generation. It aims to produce a unique graphic style characterized by:
Fine, sharp, and precise lines
Realistic and detailed characters, bridging anime and comics styles
Backgrounds balancing between realism and illustration
Usage Recommendations
For characters: DPM++ 2(3)M SDE
For backgrounds: Euler (with attenuated settings)
Use "scenery" in prompts for backgrounds
Current State and Limitations
Tendency to generate female characters, often in portrait
Over-representation of Caucasian features
Hands need improvement
Recurring patterns
Your input is not just helpful—it's essential! It will significantly speed up the update process as I work on mastering the techniques and optimizing the model. Every image you share brings us one step closer to realizing the full potential of this exciting project.
Potential and Objectives
Capable of generating known characters from video games and animation
Aims to improve through user feedback and generations
I can't thank you enough for your potential assistance. I sincerely hope you'll find the concept behind this model as thrilling as I do. So, dive in, let your creativity run wild, and most importantly—have a blast experimenting with it!
Description
LoRA to reproduce my GAMMA version of this checkpoint.
FAQ
Comments (3)
I like it! Please tell me what is the base model or models for this mix
I simply started with Pony Diffusion V6 I made an extraction of Duchaiten V7 to dose the initial change of style and merge many LoRAs (included mines) and return on the AutismMix to to get back a better prompt response (In My Opinion). I still have a lot of improvements to make, but testing the settings takes a loooot of time, so I prefer to do other things on the side and come back to it when I find a good idea to fill the gaps in this model.
I'm glad you like my work, I hope you'll like the next ones.
I just saw that it was not on the checkpoint that you asked this question. Sorry, I developed more on the base. But concerning the LoRAs, I used at least fifty to arrive at the Gamma model. From my own published LoRAs, I put PhonyFrazwell, Kittcase and a little Cyber manhwa. The rest is line sliders, detailers and some others trained with many images collected of the desired style to train LoRAs specific to the missing improvements, that's why I asked to helping me by publishing your results here to feed next LoRAs... but since success is not there, I have to generate it myself image by image, so it takes a long time.