This model is a departure from my ArmorPunk model, this model is focused on knight armor and other medieval armor, I also added some spice in the training data with monsters so you can also make some interesting armored monsters. It is recommended to use a style lora for base PonyXl if you want a specific look. If you want you can support me here: Ko-Fi
V2
Completely new and expanded dataset with a lot more tags, I recommend using a style LoRa for a consistent art style if you're using the default PonyXl model. I also added some concepts just for fun.
Usage:
Trigger word: arm0rs3nt1n3l
Strength: 0.5-1 but feel free to experiment
THESE CONCEPT WORDS ARE OPTIONAL YOU DO NOT NEED TO USE THEM! New concepts, just add them in the positive prompt, you may need to up the prompt strength:
-viking
-spartan
-templar
V1
Usage:
Trigger word: arm0rs3nt1n3l
Additional tags: armor, full body armor, shoulder armor, gauntlets, armored boots, and basically any type of armor you can think of it will work give it a try.
Strength: 0.6-1 but feel free to experiment.
Description
FAQ
Comments (4)
I like this LORA, but is there any way it can make helmets or armor without spikes all over it ? I told it smooth mirror finish and I get what looks like a pincushion
Thanks for the feedback I looked into it and when I was tagging the images I forgot to do a description for the helmet part, V2 is in the works I'll address the issue and release it soon, sorry for the inconvenience.
@NeoWave No problem at all. I love the concept, it's just frustrating. My prompt was:
score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, arm0rs3nt1n3l, in castle courtyard,
solo, 1boy, full body, impenetrable (smooth mirror finished) black and red helmet with (no sharp edges), red and black full plate armor with (no sharp edges), extremely detailed and articulated,
<lora:ArmorSentinel:1>
@TecKnight Here you go try it now if you're still getting unwanted spikes maybe put the word spikes in the negative prompt but I think it should be good now the dataset is much larger and I completely redid the tags.