Type | |
Stats | 862 2,832 |
Reviews | (121) |
Published | Nov 9, 2023 |
Base Model | |
Usage Tips | Clip Skip: 2 Strength: 0.8 |
Trigger Words | wearing linothorax armor wearing linothorax armor and feathered helmet tunic, greaves, sandals shield, spear (READ NOTES) |
Hash | AutoV2 10D1074DC8 |
Next on my line of historical outfits, and the sequel (prequel?) to my roman legionary armor.
When people think of ancient Greek warriors they usually imagine the full bronze armor version (Troy) or the half naked one (300). But in reality the lighter, cheaper and most conformable linen armor (linothorax) was the most popular choice for a couple of centuries around the Mediterranean.
WARNING: this LORA is quite random compared with the roman one, as there are simply too many variations of the armor (design, color, metal reinforcements, etc....). It has quite a good grasp of the basic shape, plus it has learned also the rest of the equipment: helmet, tunic, greaves and sandals.
Trained on base SD 1.5 (Clip Skip 2) using around 100 images combining photos and illustrations. Recommended weight of around 0.8
The base keyword is: linothorax armor
But actually the most common prompt is: wearing linothorax armor and feathered helmet
Additional keywords: tunic, greaves, sandals
Others: shield, spear, sheath
THE GOOD
Can generate a decent linothorax armor and the corresponding helmet, focused on the corinthian type
Tunic has been tagged, and if your prompt includes it you should be able to control the color. This color will also impact some elements in the armor (see samples)
Should work fine with both photos and illustrations
THE BAD
Using "fur" in the negatives could help if your checkpoint is adding it to the shoulder pads. Seems to be a side-effect of the "feathered" for the helmet...but if you remove it, you may lose the crest. Still checking it
It's SD: don't expect the leather straps to make much sense
As with the other one, almost no images of women in the dataset. It should work, but some of the shape will be lost
THE UGLY
The dataset includes a lot of images with spear and shield, but as usual SD is going to struggle with that
A shield is included in almost all the training images, but sometime its shape is not clear enough for SD. Meaning, you will get shields even when not prompted. For that (or any similar issue with the greaves/helmet) try adding it to the negatives