Sign In

AdriaticInkVision style [PonyXL]

29

211

101

14

Verified:

SafeTensor

Type

LoRA

Stats

211

101

16

Reviews

Published

Jul 18, 2024

Base Model

Pony

Training

Steps: 9,450
Epochs: 5

Usage Tips

Clip Skip: 2
Strength: 0.8

Trigger Words

(1nk:1.1), adr1at1c, (v1s1on:0.4)
1nk
adr1at1c
v1s1on

Hash

AutoV2
7D503E8FCE
default creator card background decoration
NeoWave's Avatar

NeoWave

I wanted to make a style model of the early 2000s nostalgia, specifically the Adriatic part since I'm from Slovenia, I recently rewatched Porco Rosso my favorite anime movie ever and got the idea and wanted to capture the feeling of Porco Rosso this nostalgic melancholy feeling in my own way and so this model came to be. For the training images there are 316 in total most of them are taken by me, then I added some old photos from the Slovenian public library and old paintings, then I added some old posters and some images I had laying around (anime/cartoon) that I put through img2img then upped the saturation in Photoshop. It took me a couple of weeks to gather all the images and scan them into the digital format, it takes time and resources to make these so your support is greatly appreciated, commenting and posting pictures helps a lot, thank you kind people and as always happy generating!

I love making models but sometimes it's hard managing life and free time so if you want to support me financially you can do so here: ko-fi

Usage:

There are 3 main trigger words in this model:

-1nk

-adr1at1c

-v1s1on

Recommended model is r_rain's RainPonyXL or alternatively PonyXL the results differ a lot. I tested only these two models but maybe it works good on others too try it and see.

The short version:

If you don't want to experiment with the tags just use this: (1nk:1.1), adr1at1c, (v1s1on:0.4)

and type away.

The long version:

Basically what these trigger words do each one calls upon a different dataset in the model:

-1nk-basically images are more cartoony

-adr1at1c-makes the images more colorful and is good for landscapes, also makes images softer more painterly style

-v1s1on-makes the images more realistic, also sometimes makes characters hold up peace sings at least in my testings

Feel free to experiment with these tags I usually use all of them but if you want something different try removing some or changing the strength. You can also remove all of the tags and go with no tags at all try it experiment, have fun.

For strength I use 0.8 or 1 but you can go lower if you thing the style is too heavy 0.6 is also good I think. If you're not getting the desired results feel free to message me I'll do what I can to help.