Noctis | Final Fantasy XV (SDXL Pony)

Noctis Lucis Caelum from Final Fantasy XV + Brotherhood

Brotherhood version: was trained with high quality screenshots from the anime only. No fanart, renders or 3d game engine depictions were used. The Brotherhood style is intentionally strong, as the goal was also to keep the style consistent, there are more style-flexible models of Noctis already on the site.

Official Game Model version: This was trained with high quality in game images only. No fanart, renders or other non-luminous engine depictions were used. This was trained solely on the default outfit, and in particular, used the Alternative Prince's Fatigues by Kizari to normalize the outfit a bit to help it train better with consent of the author.

This LORA will not produce usable images with models not based on Pony XL.

Activation text for the Official Game model version is just: Noctis

Tag suggestions for the Brotherhood model:

Default model
noctis, 1boy, black jacket, short sleeves, single glove,

School model

noctishs, 1boy, teenager, suit,

Younger model

noctisyoung, 1boy, young, black shirt, wheelchair, bandages, sitting

Tag Suggestions for Official Game Model:

Appearance:

noctis, 1boy, black jacket, black shirt

Outfit:

black shirt, black jacket, fingerless gloves, black pants, black footwear,

Pony seems way more trained for females than males, so the checkpoint you choose is very likely to make a large difference in results. I found some checkpoints will look great with basic prompts and basic settings, and the others either just look awful or arent as easy to work with.

Not entirely sure what the best parameters are, I haven't had any issues generating with just my basic settings:
Clip Skip - 2

CFG - 7

Steps - 60

Sampler - Dpmpp_3m_sde_gpu

Scheduler - Karras

I'm a modder for Final Fantasy XV, and an enthusiast for AI art. I will try to get the rest of the cast soon.

Support me on:

https://www.curseforge.com/members/rinual/

https://ko-fi.com/Rinual

This model is very trained, and my second time training without captions. You will not need to prompt parts of his outfit, it should default to it unless you specify something else. But stable diffusion will assume his is wearing his outfit by default. With that said, these are tags that should aid in generating parts of his outfit.

Weight: 1.0

Trigger words: noctis

Appearance:

noctis, 1boy, black jacket, black shirt

Outfit:

black shirt, black jacket, fingerless gloves, black pants, black footwear,

I generally will only ever use the prompt:

For models where it looks a bit cooked, i assume the pony model was not trained much on males, try lowering the lora weight.