Somebody commented that this is the last thing to do to close all the iconic suits from SW, so i did.
Updated with Krea2 version:
-Krea2 has great knowledge, but it does not get the uniforms correctly out of the box, even if it is close. It tends to add buttons and some shoulder patches.
-Trigger is "IMPOFF uniform", You can specify the color if needed.
Updated with LTX2 version:
-Trigger is "IMPOFF uniform", You can specify the color if needed.
-Back of the uniform is mostly not correct, can be fiixed later.
Updated with Flux version:
-Trigger is "imperial IMPOFF officer uniform", but "imperial officer uniform" should also work. You can specify the color if needed.
Trigger is imperialofficer uniform.
Updated V2:
-is much more consistent and better looking.
-You can now use "white imperialofficer uniform" to get ISB version from Andor.
-Best results with 0,7 or 0,8 weight.
-It no longer tries to add buttons and skirts.
Add "hat" to the prompt to get the headwear.
As there is no one color of the uniform you should add the desired color to prompt if you want to have a specific one.
Description
trained in 768,768. Adjusted the data set
FAQ
Comments (16)
This looks amazing. Is it possible to get the white variant to match the Grand Admiral uniform from Rebels or Legends?
You can try, but it was not trained on that variant, so the collar and shoulders will probably be different.
@impossiblebearcl4060 That's unfortunate. But I'll see what can be done. Thanks!
Wow, it looks great, i've been training lora clothes for a while, but i always have problem with face or hair color being trained also, i can't even use other celebrity loras, i only see artefacts, any advice how i can fix this issue, your model does it amazing
Only way i found to do it is to have very varied data(lots of different people) with a lot of images.
@impossiblebearcl4060 i used 50 images with 50 different faces but, it trained hair color and nose stuff, idk if captioning face will help, could you show me an example of captioning u do
@Astro9 You can look at my captions as they are in the lora metadata (edit the lora file as text).
@impossiblebearcl4060 out of curiousity when doing that (inspecting the captions in a text editor), what's the logic behind using things like "p2im" "tdrh" "tm1p" when referring to a subject in the prompts? This LoRA is one of the best I've seen by far for clothing so I'm just curious what that's doing versus referring to a male/female/etc.
@canteveninto540 I like to caption every subject with unique identifier. I noticed that it gives slightly better results. Captioning special features (hairstyle,color etc) will also give better results, but i dont like it because it takes away the flexibility of this feature,What i mean: If you have one person in the dataset with grey hair and you caption that, then after using "grey hair" in the prompt you will get something similiar to that person.
@impossiblebearcl4060 That actually makes a ton of sense, I'd have never thought about that - tysm for taking the time to answer & keep up the great work!! The sci-fi themed clothing LoRAs are deeply appreciated!
Great lora, but I'm having trouble trying to make the Togruta, Nautolan loras to work with the uniform.
Big thanks to the author! The original version was good, but the finalized version allows for just phenomenally stable results. Greatest work!
Thanks, if you have any other uniform ideas, then dont hesitate to share and i will try to make such a lora.
Question. What tags such as dmmr, t2mp, bc1t, cg2g, etc mean? I can guess that trkn means Tarkin, but what about the others?
I don't remember exactly which was which, but it doesn't really matter as those tags are not very good. Tmp means a character that was an extra and does not have a name.
@impossiblebearcl4060 ah, ok. Thanks for the lora and such a quick response!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.
















