hat feather, hat, armor, long sleeves, plume, belt, boots, smile, jewerly, (sword, holding weapon)
Description
FAQ
Comments (12)
Hi, good to see more love for the Landsknecht here on Civitai. One thing though, is this only for a single pose? Can you include poses such as swinging the zweihander, wielding a halberd, overhead swing, side-swing, guard, etc. pose in the future, if possible? It's really hard getting any particular attack pose to work
That's right. Currently, only standing poses are possible. However, complicated forms like swing a sword (especially if it's a two-handed sword) are hard to get material to make lora, and very hard to do with just prompts and lora. It might be more helpful to use ControlNet (Depth, OpenPose). However, if I get enough data and my know-how goes up, I'll try it.
Thank you.
@passinghunter1214911 Have you considered using artworks mixed with realistic photos and tagging the pose and style respectively? Something like "Warhammer Fantasy Empire Greatswords" would yield quite a bit of artworks as well as miniatures in various poses. Potentially online video for Zweihander and Montante usually has re-enactors showing off historical techniques with the appropriate sword, if not outfit?
P.S.: my review imagine is actually rendered in t2i using openpose CN and a reference CN with the same picture of an Warhammer Greatsword holding the sworld with both hands upwards at the shoulder, but no matter how much I try, even using GIMP to paint an upward sword then generate with i2i, it still shows the sword pointing down. So sadly it's very stubborn if I include the token of "sword"
@cliang96844 Miniature is not suitable for learning because the proportion of the human body is different from that of real people.
In the case of video games, it is expected that there will also be a problem with resolution.
Artwork or photography is the best way to use, but for accuracy, I have to make all the separate lora according to the shape of holding the sword, but I think it will take a lot of time and effort.
The reason why only the standing pose is possible is that the data was found the most.
If you use CN, I will test it and let you know.
@cliang96844 I added a sword-raised Landsknecht image to the gallery. I used ControlNet's Canny feature. Of course, it's not perfect, but there's also a way to implement it this way after getting the right reference image, so please refer to it.
Hi can I ask what models you used? I'm really struggling getting the lora to produce the same unique, colorful clothing.
Which image exactly are you talking about? Two types of models were used.
@passinghunter1214911 https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/108b82e8-7202-4027-be75-4cc0c635ac1b/width=450/00057-925860179.jpeg this one
Do you consider remaking this model for Pony?
I'll try