Based on the popular Pop Up Parade figures!
Inspired by this LoRA, check it out! There's also another Pop Up Parade style LoRA from the same dump as the figma one here. (Also a nendoroid one)
But I already started collecting images when I came across it, so I tried training it myself to see how it turns out.
If you want to support me buy me a coffee!
Weight: 0.7-1
Triggerwords: None.
Notes:
Results are very dependant on the checkpoint/model used. I used AOM3.
Feel free to share some of the pictures you create with me and post them here!
Description
FAQ
Comments (15)
I never knew i would need this...
Glad you like it!
It's really fun to see Pop Up Parade figures of characters that are way too niche/unpopular to get turned into an actual figure, that was part of the reason I did it.
(Big fan of your Rangiku btw 👀)
I can see this LoRA being very useful to merge into a base model at low weight to give it a distinct style. In fact, I just may do that myself.
How do you scale once you have the base generation to make it into a cleaner figure?
I used Latent as upscaler, upscaled by 1.5 (base resolution set to 512x768) with the Denoising strength set to 0.5-0.6 for most of the example pics. 0.5 on the one you commented specifically.
I am however not sure if those are ideal settings.
I usually mostly do anime pics for which I use R-ESRGAN 4x+ Anime 6B, figures however border more on the photorealistic so I am not too experienced with that either.
@Beelze can I add you on discord and show you the current error I'm getting?
Rasuki#8619 - here is my discord, hopefully I can show you the issue in DM, it may be a lora clash
@Rasuki I'm not sure if I can help you with that :/ I'm pretty new to this whole thing myself! Maybe just post the error code/message here so people can help you out!
Or general information like what model you are using, what other LoRAs you are using if any et cetera!
I'm sorry you are having issues, hopefully someone can help you if I can't.
@Beelze I'm using the same settings as your outputs but with the aqua konosuba lora but I can't seen to get the upscale right, just keeps ending up flat
No worries if you can't add me, would just be super useful to share processes and see why the change is occuring
@Rasuki
Flat in which way? :o
The only way I manage to make it flat is by using a model that is mostly 2D based and can't handle the figure aspect, Anything V5 for example tries to shade it like a drawing which could make it look ''flat'' even though it is very versatile usually. So for this LoRA I usually use AOM3.
like this https://files.catbox.moe/xm13r0.png
I just downloaded all 3 of the Aqua LoRAs (by Lykon, Pleased_Chomusuke, and Pikkiwynn) I found and will try to make images with them to see if the issue is with the Pop Up Parade LoRA or somewhere else at least!
I'ma post the images as a post below (will change the Lora names to put the author at the beginning and just use the activation tags and quality tags, so you know which is which.
They all work (more or less) fine, some tend to struggle more than others. Maybe it's the checkpoint you are using? I am really out of ideas here if you aren't x)
@Beelze yh it's been super weird, I'm still playing with it now, the body seems to conform when I move the lora for figures to the front of the prompt but the face just always seems 2D
what are the right promps for making anime figures?.