Not Artists Styles for Pony Diffusion V6 XL

Not Artists Styles for Pony Diffusion V6 XL - Summer Night

NSFW

Styles are trained on synthetic data and do not copy the styles of artists!

Photo 2 - is trained on real data (Dataset from Unsplash)

[M] Merged LoRAs

Recommended LoRA Strength (Weight): 0.8-1

Preview generated without using ADetailer & Hires. fix

Description

Repeats: 20

Epoch: 20

Image Count: 16

Network Rank: 32

Network Alpha: 16

Batch size: 2

Clip skip: 1

Training steps: 3200

Resolution: 960x1024

Train Resolution: 1024x1024

Model: ponyDiffusionV6XL_v6

WD14 Captioning: SmilingWolf/wd-v1-4-moat-tagger-v2

Trigger Words: dark theme, low light,

LoRA Weight: [0.8-1]

With the use of Trigger Words the image becomes darker and more like night, same with Cold Night

FAQ

Comments (13)

TechnoyoteMar 4, 2024· 4 reactions

CivitAI

Just wanted to say I think your recent models have been your best yet. Thanks for the awesome tools!

heathergreen95Mar 4, 2024· 4 reactions

CivitAI

YES! Thank you for updating Summer Days to V2! This is THE BEST anime style for Pony!!! Also, you fixed SDXL's issues with night/day scenes... simply amazing

Rating_AgentMar 4, 2024

CivitAI

Please new version of Gothic! Its a best one!

1376925Mar 6, 2024

CivitAI

Is there a specific method to training for Pony? Or you just use it as the model when training and all else is the same?

obuyb404

Author

Mar 6, 2024· 1 reaction

I don't have any specific method of training models, just some SDXL Preset I can't remember which one anymore, which I changed a bit and that's it

I saved my Preset for Kohya_ss and use it pretty much unchanged

ㅤ

About training the Pony XL model, this model doesn't require a large dataset.

For style training only 10-20 images will suffice, And for character training only 20-40 images will suffice

The bigger the dataset the better, but you can't always find a big dataset.

ㅤ

This style is trained on only 8 images.

Smooth Style

https://civitai.com/models/264290?modelVersionId=297988

ㅤ

This character is trained on only 11 images.

Ferdinand Clooney (Arknights) LoRA

https://civitai.com/models/320971/ferdinand-clooney-arknights-lora

PangolinMasterMar 6, 2024· 1 reaction

CivitAI

Any plans for a smooth anime night?

obuyb404

Author

Mar 6, 2024· 1 reaction

Yes! Only Smooth Anime consists of two models one model is already Cold Night (Night Version of Anime Style).

Smooth + Anime 50/50 = Smooth Anime

DreamImgMar 6, 2024

@prgfrg23 I'm sorry but I'm confused by your sentence. Are you saying that Smooth Anime already has Cold Night baked into it and can use the night trigger words? Or are you going to make a new version that is Smooth Anime Night?

PangolinMasterMar 6, 2024

@fatcan I was a bit confused too but I assumed he means cold night is an anime style

obuyb404

Author

Mar 7, 2024· 1 reaction

No, I didn't mean that Smooth Anime has Cold Night (Smooth Anime doesn't have Cold Night)

To make a Night Version of Smooth Anime I need to merge Smooth Night (Doesn't exist yet) and Cold Night.

Smooth + Anime (Not Cold Night) 50/50 = Smooth Anime.

Smooth Night + Cold Night 50/50 = Smooth Anime Night.

I hope that the name Cold Oil/Cold Oil Gothic is not misleading, but now I'm not sure (Cold is just a name that has nothing to do with night).

Sorry if I confused you, I didn't mean to.

One part of the model already exists is Cold Night but to create Smooth Anime Night I need Smooth Night (This model doesn't exist yet).