π΄Patreon | Choose what i'll train next | PANyZHAL
ββββββββββββ β¦ β β¦ ββββββββββββ
weight: 2.Β always. ALWAYS.
Hi! I know I only train this, but I like it so much! I hope you enjoy it too. This LoRA is supposed to work with models trained on SDXL because pony models aren't too good with backgrounds and SDXL LoRAs can't work on them. So, this model opens the possibilities to a new world.
Actually, the pony LoRA can also work with other SDXL models using the same trick that I am going to teach for this model.
The trick:
Put the
weight: 2.Β always. ALWAYS.
With this, in some generations, you almost don't need keywords, but to be sure, use them: disembodied head, headless, detached.
Recommendations:
WEIGHT: 2
For complex poses you can use depth maps

Description
You need Weight: 2 on this LoRA.
It's still not perfect. I didn't really add more data, just pruned some tags and made it for SDXL models. I'd love to make it work for complex poses. Anyways, I made a lot of attempts without success, but since these models work on SDXL and ControlNet works better there, you can create better poses with a depth map.(see the example)
FAQ
Comments (9)
A tip. Add the following negative prompts: "doll, mannequin, fake head, plastic," otherwise the disembodied head may look like one of those things. (Unless that's what you want).
sd version please!
hi! sorry for the late reply, what you mean by sd? if you want sd 1.5 i already have one, it's more difficult but with some effort you can generate good stuff
Disembodied Head/Headless | NOW FREE - Headless/holding head
don't skip this lora just because it's older or whatever, it does some crazy shit when you combine it with the other dullahan loras. None of them work well on their own but you throw this in and all the problems go away
I concur. Will be eagerly awaiting the day a flux version of this lora comes out. I can already do good Disembodied Head stuff with Flux raw. But Dullahan/Holding Head stuff is certainly Lora necessary.
@adminloliconrocks Hi! Thank you for your kind words! I'm really happy to read that the model works for you. It can be tricky, but I really put effort into this concept, and knowing that it works even better in combination with other models is good to read!
@tunneltube951Hi! Iβve been out of the training game for a few months, but now I want to get back into it. I havenβt really looked into how to train FLUX Loras yet, I just know the requirements are pretty high. But if I find a way to do it, you can bet I'll be making a Lora soon. For now, a buddy of mine got ahead of me, so you can give his a try.
Headless - Flux - v1.0 | Flux LoRA | Civitai
you know i was trying to build a dataset for this model but it really caused a problem with the local police... thank you for making this lora
I love it
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.











