Update For T2V
I trained T2V using the same dataset as I2V and achieved OK result. The lactation part doesn't work consistently but otherwise acceptable. I didn't do a lot of testing and please feel free to share results and feedback
General Guide
I have been obsessed with lora training and this is my third model in 3 consecutive days despite my busy working schedule. I'm trying to quickly build out more basic NSFW motion/concept models for the degenerate community, so fellow content creators can pump up interesting work and make me laugh, lol. Training T2V and maybe writing up some lora training guide is on my roadmap, stay tuned.
You would want to use prompts like this:
One person in the middle of the screen, Her hands cradle the breasts and she milk her breasts by squeezing her breast nipples. loads of jets of milk shoot from her nipples.The keywords are
squeeze, massage, nipple, breast, jet of milk, droplet of milk, leak/shoot from nipplesAs usual my lora works between 0.8 ~ 1 strength. Combining it with my other lora like sexmachine will give good results. I've also tried ahegao lora and it kinda works. Feel free to explore more combinations.
Description
Initial version
FAQ
Comments (21)
good work bro. Please do a creampie Lora, with & without penis inside pussy !!!
thanks for the suggestion, will consider
Your job excellent as always. Love it >_<
god bless you!!
U doing gods work
Man slow it down you scare me
@shura490 Sorry your Lora works to well :P
I'm getting the following error while loading the lora
Loading Musubi Tuner format LoRA...
lora key not loaded: diffusion_model.blocks.0.cross.attn_k.lora_A.weight
lora key not loaded: diffusion_model.blocks.0.cross.attn_k.lora_B.weight
lora key not loaded: diffusion_model.blocks.0.cross.attn_k.img.lora_A.weight
lora key not loaded: diffusion_model.blocks.0.cross.attn_k.img.lora_B.weight
lora key not loaded: diffusion_model.blocks.0.cross.attn_o.lora_A.weight
lora key not loaded: diffusion_model.blocks.0.cross.attn_o.lora_B.weight
lora key not loaded: diffusion_model.blocks.0.cross.attn_q.lora_A.weight
lora key not loaded: diffusion_model.blocks.0.cross.attn_q.lora_B.weight
lora key not loaded: diffusion_model.blocks.0.cross.attn_v.lora_A.weight
lora key not loaded: diffusion_model.blocks.0.cross.attn_v.lora_B.weight
lora key not loaded: diffusion_model.blocks.0.cross.attn_v.img.lora_A.weight
lora key not loaded: diffusion_model.blocks.0.cross.attn_v.img.lora_B.weight...
100 more lines here.
What am I doing wrong? I'm using HunyuanVideoLoraLoader node. All other wan loras work fine
Are you trying to use hunyuan model with this lora, does it work? I have no experience with Hunyuan. I think the error is pretty literal - your loader is expecting a certain format of file but mine isn't that format. I can try to port my model to Hunyuan at some point, otherwise I can't really help as I'm not a low level expert of Lora.
No no, I'm using Wan2.1 model. The HunyuanVideoLoraLoader works with Wan2.1 loras as well. This is the first lora I faced this problem. All others wan loras worked without giving this error.
@honryindian have you tried loading it with a normal, native lora loader?
@honryindian If you publish your file I can take a look
@coyotte followed your suggestion and now it works! Thank you!
@shura490 Got it working with native lora loader. Thank you for the amazing lora!
@shura490 The native lora loader with clip has no effect I have tested it Wan has a very good native handling of xxx thats what I have discovered. Pls try to provide the Loras in diffuser format.
@shura490 I have converted the lora and now testing lora was loaded without ban error: [DisTorch] Full allocation string: #cuda:0;22.0;cpu
Loading Diffusers format LoRA...
Loading Diffusers format LoRA...
Loading Diffusers format LoRA...
As for conversion, command is python convert_lora.py --input path/to/musubi_lora.safetensors --output path/to/another_format.safetensors --target other its in the repo only the convert.py and the util folder is necessary https://github.com/kohya-ss/musubi-tuner
Nice! How about a self-suck-nipples-lora?
will consider
@shura490 Great work on these concepts, thanks!
If you're open to ideas i think a modeling to sex lora would be really useful?
Like imagine a soft to hard porn concept where a guy comes from off screen into the usual woman standing pose images. The guy would appear and pull her skirt/panties down and have standing style sex. This would allow us to turn any of the countless model images into proper sex scenes easily.
If it works well with standing perhaps it could be improved to allow more elaborate poses like bent over etc. I think to begin though trying to mostly preserve the standing pose would be good for consistency and coherence. I suspect if the woman is moved too much it might cause distortions to face etc.
I'm not sure how tricky this would be to do but after seeing all the fun loras where someone comes into hug or a scene completely transforms i think it might be really cool to have?
@sdimg Think this will be better split into two to three different loras and use clip extension tricks to realize. The first one would be a person walks into the scene, and then taking off pants and having sex will be their own loras. Otherwise I'm gonna need endless time to train one lora since the concept is too complicated.
@shura490 Yeah it could well be tricky! Thanks for taking a look and considering it, hopefully its doable!
Details
Files
lactate-shura-000002.safetensors
Mirrors
lactate-shura-000002.safetensors
lactate-shura-000002.safetensors
116_lactate-shura.safetensors
lactate-shura-000002.safetensors
lactate-shura-000002.safetensors
wan_14b_milk_lora.safetensors
lactate-shura-000002.safetensors
lactate-shura-000002.safetensors
lactate-shura-000002.safetensors
Wan22_Tt_lactate-shura-000002.safetensors
lactate-shura-000002.safetensors
lactate-shura-000002.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.