Trained to reproduce ongoing unbirthing scenes.
It should work with a wide range of settings, I suggest strength 0.9 but others are fine too.
The base key prompt is "unbirth".
You can add these to influence the result:
"unb_up, upper body out from vagina" for the upper body of the unbirthed character coming out.
"unb_low, lower body out from vagina" for the lower body and legs coming out from the vagina.
This is not guaranteed that it will work as intended, and will sometime produce the opposite situation.
Description
First version
FAQ
Comments (27)
This is a neat lora, but I'm having a really difficult time trying to get the legs sticking put, normally it's head is out or only head is in. Very finicky.
Not gonna lie, this is the best birth lora I've worked with, conventional or unbirth. Would it be possible for you to make a lora for standard and breech births of normal offspring?
I know it's greedy AF to hope that this gets worked on more... it's pretty good and it's the concept I've been waiting and hoping for, but it's quite volatile and tricky to control in its' current form.
Super-appreciated regardless!
Could you please add Lora "Man's head inside vagina" ? Just like woman sucked man or boy into her vagina.
Sometimes ((head in vagina)) , or ((head coming out of vagina)), works.
I like your work but, thanks to Civitai's inept handling of aliases, I can't find "lora:Realistic000003_1_2.5DAnime_0.05_Merge1", which you and others often use. Please, give me the Civitai link for this lora or its alias that works in Civitai's search engine.
@nizur i can't find it either... May be author changed the name of it? i can send you a file, if you want. Description in my A1111 is:
ss_base_model_versions
dxl_base_v10
ss_mixed_precision
fp16
ss_network_alpha
16.0
ss_network_dim
32
ss_network_module
networks.lora
ss_output_name
Realistic000003_1_2.5DAnime_0.05_Merge1
ss_sd_model_name
ponyDiffusionV6XL_v6StartWithThisOne.safetensors
ss_v2
Falsesshs_legacy_hash64df3ab8
sshs_model_hash
1a85c6cf999667f7f64fa3319d7799e2f09107a304ab5c910ccb404ebf0fe7c2
Thanks. I asked on Reddit and somebody found it. It's https://civitai.com/models/297619?modelVersionId=349828 . I have a tip for you. Experiment with dynamic lora weights. For example, <lora:Unbirth: 1.5@0, 1.5@9, 0.8@10>.
@nizur Thanks! Is there any manual for using dynamics lora weights?
https://github.com/cheald/sd-webui-loractl
I assume that you use Auto1111. You have to install this extension.
@Bugagish
https://github.com/cheald/sd-webui-loractl
I assume that you use Auto1111. You have to install this extension.
So might as well ask here, but I'm in dire need of examples using BREAK, I think that might help with getting the head out of vagina to show better especially with 2 different characters. (2 women)
@Taco360 I am not an expert using BREAK but I doubt that it will help you much because this unbirth lora follows prompt instructions very poorly and many images have to be generated to get an image that matches the prompt. If you examine the information on the renbaum images (the ones with latex) you will see that he incorrectly used unb_up when he should have used unb_low but still gets what he wants, probably because he generated a lot of images.
Great job! May be in future we'll see LoRa with some hands on the bodies, that inserted or shoved. I mean, they literally shove the victim's body into themselves with their hands.
Can you please do some non hairy examples of 2 women?
@Taco360 i'll try)
what i would give for a 1.5 version
It's not a concept possible on 1.5 so the answer is no sry.
@JollyIm it is just extremely difficult, only the furry models/hll can kinda pull it off currently
@nova1337 yea sorry I mean it's impossible with the dataset used for doing it on pony, and I'm not going to put work to make it for 1.5 that I'm sincerly not interested in anymore. It's incredibly tedious to do concepts like these for 1.5.
@JollyIm understandable, thanks for replying anyway, once i can get a better gpu will give ur loras a try :D
hi every one
Can you please tell me what this is (pony)
I download and install "ponyDiffusionV6X" in stable-diffusion-webui as checkpoint and use this file as lora, but nothing happens. What is my mistake?
unb_low doesn't really influence the results, but still it is a good LORA thanks so much for making it :)
plans for illustrious?
hopefully one day, illust 0.1 ver
how the hell does one train this thing. based on what sort of pictures.
On like, thousands of images in imageboards, art sites, e621, rule34, aryion, danbooru and thousands of others? What kind of question is this lmaoo
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.