Fairy Tail Girl Pack
After a commission ask - here the new 0.4 Update
Samples and Grids in Gallery
- Its alot of Work big Character Packs,if you want to Support me see my Patreon. [Commission open] -
Roadmap:
Guild Tattoo
Better Train Image Rework
Stable Gen / No Overtrain
Changelog 0.4
Adding 6 new Girls
Meredy
Virgo
Aries
Aquarius
Sorano
Belserion
Rework
Cana and Evergreen full Rework
New LoRA Config
Bugs: If you get "overtrain" Images change weight to 0.4 and/or cfg to 4 fix many times my images - See grids.
This will be fixed on the Road to 1.0 :) (yes its still beta lora :D - Flexible big packs hard work)
Version 0.4 Girl List:
aquarius
aries
bisca
brandish
cana
erza
evergreen
flare
hisui
irene
juvia
kagura
levy
lisanna
lucy
mavis
meredy
millianna
mirajane
sherria
sorano
virgo
wendy
yukino
Full Version 0.3 Girl-List
bisca
brandish
cana
erza
evergreen
flare
hisui
juvia
levy
lissana
lucy
minerva
mirajane
sherria
wendy
yukino
Version 0.2 NEW Girls [WiP]
Kagura
Mavis
Milianna
Minerva
Sherria
+Version 0.1 Girls
Version 0.1 Girls:
Bisca
Cana
Erza
Hisui
Juvia
Levy
Lissana
Lucy
Mirajane
Wendy
Yukino
Trained on: W:480 H:720
Triggerwords:
"name" - see list
fairy tail - Trigger for girl name
anime art style - Trigger for Anime Style (Beta)
See Preview Images for my Prompts.
Preview Images generate:
LoRA Load over: additional-Networks (work without it)
Model: Anything 4.5
Txt2Img and Upscale settings:
DPM++ 2M SDE Karras
Steps: 20
Sample Size: 512x768
CFG: 4/7
LoRA Strengh: 0.4/0.6
Upscale: 2x
Denoise: 0.6
Upscaler: R-ESRGAN 4x+ Anime6B
Description
LoRA Strengh 0.6 + Highres Fix recommend
More Flexibility
FAQ
Comments (21)
add Brandish
@shekelberg the 0.3 is in make :)
what is the best program to use this with ?
I use WebUi + additional-networks
NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.
Time taken: 0.32s
Torch active/reserved: 2391/2594 MiB, Sys VRAM: 3941/15102 MiB (26.1%)
How can i fix it?
start webui with --no-half or --disable-nan-check
or use other VAE or other prompts
@sysdeep I got same problem
Can you tell me where exactly to put --no-half and disable-nan-check?
I put it in normal prompt after description above negative prompt, but it didnt work.
I use google colab to run Stable diffusion
I am really sorry, I am still new to this 😓
@DarkneelBuster Sry i dont know about SD and Colab :/ i generate all over my own PC :( hope anyone else can help you - Or ask on Civitai Question Area
@DarkneelBuster Probably you are doing what I did. I was using the Lora in the wrong place because I didn't know anything. Please see some videos, you'll get what I mean :D. Btw you don't need to change any configuration.
Sysdeep I did a few using AbyssOrangeMix2_hard.safetensors and the Lora but the quality was really poor. Was it because of steps or any other configuration?
@15501550zzz1550843 All my Generation Config are on Bottom of my Model-Pages - steps / cfg / highres and more
@sysdeep probably I doing something wrong because I copied your config. Maybe is the model wrong....is the same the model and the Stable Diffusion checkpoint?
@SysDeep yups.. I am using colab because I don't have good PC 😔
Okay, I will try to ask in civit forum or Stable Diffusion discord later.. thanks
@15501550zzz1550843 I've tried using AbyssOrangeMix2_hard.safetensors too before..
it's work but same like you too, the result is different unlike SysDeep did.. it became chara illustration style. Not like anime style
I've checked some YouTube videos but still can't find the solution for colab user 😭
Have you found any solution?
@DarkneelBuster I Installed SD complete new on my Second PC and i can 1:1 recreate my Images :/ i dont know what is wrong
Hi, Did these LORA have guild tatoo prompt?
Not in v 0.1 + 0.2 its come to road to perfect LoRA :D (1.0)
add flare corona plsss❤️❤️❤️
Done :) check 0.3
Hi, thank you for the great LoRA. Can you train for Brandish also?
Done :) check 0.3
@SysDeep appreciate your hard work.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.