Hu Tao LoRA Model Card
Overview
This model captures the distinctive and vibrant aesthetics of Hu Tao, the beloved Pyro character from Genshin Impact. Hu Tao is known for her playful antics and eccentricity, which belies her serious role as the 77th Director of the Wangsheng Funeral Parlor. This LoRA (Low Rank Adaptation) translates her unique personality and fiery abilities into stunning visual interpretations, perfect for fans and artists looking to incorporate her character into their AI-generated artwork.
Recipe:
1 cup sass
3 cups WHOOPS.
5 cups WHERES THE. CHEETOS?
Bake at 450 for 24 hours and then call the nearest fire department.
Read a few memes while you're in the ER for severe burns
(KIDDING.)
Contact Us:
Our Discord: https://discord.gg/HhBSvM9gBY
Earth & Dusk Media https://discord.gg/5t2kYxt7An
Backups: https://huggingface.co/EarthnDusk
Send a Pizza: https://ko-fi.com/duskfallcrew/
WE ARE PROUDLY SUPPORTED BY: https://yodayo.com/ / https://moescape.ai/
JOIN OUR DA GROUP: https://www.deviantart.com/diffusionai
JOIN OUR SUBREDDIT: https://www.reddit.com/r/earthndusk/
Description
FAQ
Comments (10)
Interesting choice of choosing clip 1. the pony model was trained on clip 2. Not sure how much it will effect the result.
SDXL does not work at all with clip skip not at 2, so the setting is irrelevant. (All interfaces that support sdxl should automatically set it to 2 regardless of user input, including lora training tools)
@wewewew Funny, beause this gets argued all the time LOL - I trained it on 1, because most of my peers say 1, however SDXL ITSELF was trained on 2 from what I hear - It doesn't really matter, it's sort of a "DO WHAT YOU FEEL" XD frankly last time i was in A111, it refused to even change my settings and duplicated faces so bad that i was about to give up training loras XD
@duskfallcrew It's not just that SDXL was trained on clip skip 2, it requires 2. In A111, kohya gui, etc, the clip skip parameter when using sdxl model is set to 2 automatically no matter what you put on the UI.
It's a shame that this information is still not widely known, causing so many people to waste gpu time to test training with different clip skip values.
@wewewew I think someone told me that the settings are partically due to SMALL CLIP in SDXL being Clip skip one like SD 1.5 -but that BIG CLIP is 2, but that in Kohya code it's frozen at 2, but no matter if youhave a setting to change it it just sticks to 2 xD
I tried the lora. It is able to make nsfw and sfw art. most lora I tried fail at making the tail coat from behind but this lora is decent at it.
I did actually for once take a bat for the team, did a gelbooru run, didn't wanna look at half of what it gave me XD but i tagged it, i trained it - and VOILA! :) yea - Also if my gallery hides any pics it's not because i don't LIKE NSFW stuff i have CPTSD and prefer to keep things PG-13ish in my galleries <3
Note:
As for ONSITE Generation, it'll work well with my Virtual Diffusion, but NOT STYGIAN MIX. Stygian mix is waiting for an update, and i'm still tossing up on ifi 'm keeping that on the generator and moving back to Helllaine or not.
:) Plenty of Pony XL models to choose from, and i'm sure the dataset will work ok if anyone wanted and SDXL one.
I'd appreciate it if you'd make the LoRAs smaller. Most other SDXL LoRA creators can make them with 26-52 MB. 200 MB is quite a lot when you have many of them and are even forced to store them on a NAS
Fair enough, i don't QUITE know how to do it, but i'm happy to look into it
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.