Kim Possible Pack - [Concept + Characters] Beta v1
Kim Possible Concept Pack - With simple Char Triggers! On Beta Stage - Works well but need more Finetune :)
Base Infos: Illustrious + Anima
Trigger: sysdeep_kppack
Train Image Size: 1024x1024 , 1216x832, 832x1216
Prompting: sysdeep_kppack, kp_style, <charname>, 1girl, solo?
Style: kp_style
Other Tags: pointy breasts
When you use other Style loras - remove kp_style with the trigger of your style lora trigger.
Anima Support Tags + Text
Characters-Trigger:
ann
barkin
bonnie
drakken
james
jim
josh
kim
monique
ned
other
ron
ruffus
shego
tim
wade
yori
Description
FAQ
Comments (13)
This is absolutely fantastic! I can't wait for the full update!
Keep up the packs and updates, they're amazing!
I loved Lora and I really liked the beta in Anima. Everything is excellent and wonderful. Keep it up!
However, Kim and Jim's characters sometimes look strange, or they don't have legs or anything like that when I put them with other characters. Could you please try to fix them in future updates?
Thats the reason why i call it Beta ^^ - Beta v2 will increase the dataset from 2k Images to 5k + more Characters and more Body
what checkpoint are you using for the sample pics?
Did you train Barkin on the pilot or something? He has that awful blonde hair from the first episode. In the next version hope he has his normal dark brown hair.
@ImagineBreaker barkin dont Trainer well because only 22 Images in Beta dataset - next Version will all chars Same Count of Images. Missing Charakter will come too.
One more question did you train in any of the iconic backgrounds like bueno nacho, or draken's lair, or the school or anything?
@ImagineBreaker No, location dont have own trigger. its hard enought to "simple" chars-tags with multi char support in one image in illustious ^^ - so no background
This is fantastic! I really liked the pack.
Will the second update be coming soon? And will you add other characters like Joss Possible?
Anyway, keep it up! The pack is amazing. I hope to learn how to make packs like this in the future.
Joss is in planing, Team Possible in Planning and more - Dataset creation sucks hard :D for next version i double the Images (and new chars) so i has then 5k images in the dataset ^^
This is absolutely fantastic and works incredibly well. I hope you continue with the packs.
However, there are two issues. I don't know if it's because it's a beta version or if I'm writing the prompt incorrectly, but I'll mention them in case you can fix them in the full release. When I set Lora to a different style, even after removing the kp_style and reducing Lora's appearance, it still uses Kim's style and ignores Lora's style.
Also, for some reason, no matter how hard I try to make a character look to the side or behind, it always seems to look straight ahead, towards the viewer.
These are two drawbacks (I'm not criticizing, by the way; I really like Lora, and I'm just mentioning these drawbacks in the hope that they'll be fixed).
Good luck, and keep up the great work because Lora is truly excellent.
Prompt order right?
<sysdeep_kppack>, <style trigger>, <chars>, <normal tags>
this model trained on multiple Styles (18) so he "know" what the style is ^^ and in my test i can create other style (mostly KPPack strengh 1 and Style LoRA at 1.2-2) when you change the Hair or something or generate other Char, use the Character tag "Other" - and yes hard to use this size of multi-pack lora ^^
And the "looking behind" Tags will be fixed in Beta 2