Sign In

Aketons & Gambisons (4-in-1) (Final Fantasy XI)

3

Verified:

SafeTensor

Type

LoRA

Stats

64

3

Reviews

Published

Feb 5, 2026

Base Model

Illustrious

Training

Steps: 1,140
Epochs: 15

Usage Tips

Clip Skip: 2

Trigger Words

aketon
kingdom
republic
federation
gambison

Hash

AutoV2
CF87F071DE

Created on Civitai

!! ATTENTION !! 重要 !! ACHTUNG !!

Are you an FFXI player? Do you have cool armour? Are you able to take high-quality screenshots?

Message me please! I want to immortalize FFXI's wonderful outfit designs.


This is a 4-in-1 clothing lora featuring the three national aketons and their plain brown cousin, the gambison. I always thought these things were fashionable as hell.

The general trigger word for all styles is aketon. In addition, each of the outfits has their own trigger word: gambison (plain), federation (Windurst), republic (Bastok) and kingdom (San d'Oria).

General

aketon, asymmetrical clothes, layered sleeves, buttons, double-breasted, collared shirt, belt, armband, knee boots, pants, pants tucked in,

San d'Oria

kingdom, red clothes, black clothes, gold trim, 

Bastok

republic, dark blue clothes, black clothes, gold trim, 

Windurst

federation, green clothes, black clothes, gold trim, 

Gambison

gambison, brown clothes, burgundy trim, 

Hat

[color] beret, hat feather,

More options

thighhighs, bare thighs, zettai ryouiki,
aiguillettes, gold aiguillettes (inconsistent)

Other

  • Weight of ~0.8 is good, varies by checkpoint

  • Since the lora is heavily trained on screenshots, you might wish to put 3d, realistic in the negatives

  • The lora's quite stable without upscaling, though with full body shots I always recommend a face detailer


Since there are very few decent-quality screenshots of this armour, I was forced to do the following:

  • Collect as many high-resolution-as-possible screenshots from the game as I could

  • Upscaled the screens using RealESRGAN-4x

  • Asked Sora and Nano Banana Pro to create "concept art" based on the screens

  • Added a few pieces of (only very accurate!) fanart -- fidelity is the goal here

... resulting in about 30 training images. I set the repeats, epochs, dim, alpha a bit higher than usual to compensate, hence the lora is a bit bigger than I usually make. Otherwise I think it worked pretty well.


If you enjoyed this lora, please consider leaving a gen or two below!

Thanks!