LoCon trained on Krekkov's artstyle for ponydiff: https://civitai.com/models/257749
Artist: https://x.com/krekkov | https://www.pixiv.net/users/69447461
Because of my lack of experience and time, the version number may be based on changes that have made certain aspects worse. As such a previous version may perform better than a later version.
I'm still learning the ins and outs of the model and LoRA training in general, please leave reproducible feedback or tips so that I may improve over time.
Check the "About this version" section for any quirks or pitfalls I'm aware of or have experienced.
Some Q's:
What's your workflow?
The scheduler used is Align Your Steps
Initial resolution is selected from the SDXL whitepaper appendix I: https://arxiv.org/pdf/2307.01952
Perturbed Attention Guidance is used for the initial image
Upscaler is usually 4x_fatal_Anime_500000_G or 4x_NMKD-Siax_200k
Second sample pass is the same as the first (params are on civit's meta interface) but at ~.35-.45 denoise and using the Karras scheduler
A facedetailer is used
I use ComfyUI, sorry
What strength should I be using?
Base Pony: ~.7 - 1, much more accurate adherence to style but more volatile
Autismmix: ~.8 - 1
Other Pony merges: Depends on the model and its lora compatability, I've noticed some models have really bad compat and lead to serious artifacting. This could just be that I'm bad but I notice it on other people's trainings as well so it might be a merge issue.
Sample images Model?
All images are Autismmix_confetti
What does your filename structure mean?
It's split into 4 parts of [XX]-MAnon-[YYYY]-[ZZZZZZ]
[XX] - Artist initials
MAnon - My moniker
[YYYY] - SD version and Model version (Ex. XLV4.1 is on SDXL [Not 1.5] and it's version 4.1 of the model)
[ZZZZZZ] - The epochs, only useful to me