Back by popular demand, my WAN 2.2 Bulge Animator is back with Version 2—trained for another 8+ hours on a dataset twice the size of the Version 1 dataset.
This LoRA allows your subject to rub their bulge through their clothing. It works best when the starting image already has a prominent bulge. You can create one using one of my image-model LoRAs for Krea 2, Boogu Image Edit, Mage Flow Edit, Qwen Image Edit 2511, or FLUX.2 Klein 9B.
Version 2 is better at keeping the action focused on the bulge instead of drifting up to the belly. It may still take several attempts to get exactly what you want, but it should require fewer attempts than Version 1. All of my Version 2 showcases were first results and were not cherry-picked, although I upscaled a couple of them.
This works well for foreplay and teasing scenes, especially with longer frame counts where looping movement is not a problem. It can help fill time in longer videos while teasing the viewer until your subject finally whips it out.
Without this LoRA, good luck getting the bulge to last through the full five-second video without dissolving—and you’re definitely not going to get your subject to rub it. Thanks, Coach!
Download HIGH and LOW files
This is a WAN 2.2 image-to-video LoRA that includes both high-noise and low-noise loras in the models section here. Both should be used in your WAN 2.2 workflow. If you are using a WAN 2.1-based model, my penis LoRA works with using the low-noise safetensors file, and I assume this one will act the same way if you use it with SCAIL-2 or Magic WAN, for example.
You can guide the video with either an image-to-video or first/last-frame workflow. Since this is an image-to-video LoRA, starting with a visible bulge will produce the best results. When combining it with other LoRAs, reducing its strength is recommended. You can also lower the strength if you want a less dramatic effect.
Please share your creations by clicking Add Post below. Reviews, results, and suggestions are appreciated.
Visit coachbate.ai to follow me for free to see content featuring all my LoRAs in action.
Description
You need both high and low noise lora files for this to work as I designed it. See the required files section for the other safetensors file
FAQ
Comments (41)
Tested and approved !! Amazing work, man !! :D
love your stuff. how did you create your videos? it says SD 1.5 ???
@coachbate Haha, It's because i often use the homosmile checkpoint to create the first frame, then i create the video with comfyui. : )
Great. thanks! Any idea of klein lora that can create an image with an "underboner"?
No, but here's some for ZIT and Flux 1 that I recommend:
https://civitai.com/models/2218392/erect-penis-bulge
https://civitai.com/models/1279737/vpl-visible-penis-line
and this may be the best one that just came out for Z Image Base https://civitai.com/models/2218392/erect-penis-bulge?modelVersionId=2783227
My Qwen Image Edit 2511 will do what you asked about. Just released an update for it too. See examples
I released my Klein one to do this the other day
Doesn't create the boner right? Haven't had luck in that.
Can you make a "remove singlet" lora?? Its impossible! IA makes it look like such a difficult task.
It does not create it. You'll need an image bulge lora to create the first frame image.
This worked for me on WAN 2.2 using a penis lora to take off his singlet. I'll post example: https://civitai.com/posts/27840047 - the workflow should be in the mp4 files
the man lowers his Adidas compression shorts just enough for his meaty circumcised penis to flop back and forth as moves his hips. he plays with his heavy, flaccid penis flopping it back and forth
The new Qwen Image Edit 2511 lora I just posted will create the boner
@coachbate dang I hope someone actually makes a boner lora in any ai video generator, from being normal ones to huge boners would actully be great!
@menfq77653 text to video like that for something the base model doesn't know anything about is always going to be risky from my experience. You can control exactly what you're going to get by making a first frame and last frame image. My Qwen loras for penis and underboner give good results almost every time. I run them in batch of 4 at a time, tweak or even can edit the output produced putting them as your new input image.
@menfq77653 there are some all have been deleted from here and are on civitai archive site. search there for wan 2.1 penis reveal or penis arousal - they work in WAN 2.2 also
Please, make a FluxKlein9B verison, I love it!
I create3d one but something is wrong with the site and won't let me save it. I've uploaded it with showcase images 5 times now and fails at the end.
EDIT : This site finally worked and it is posted here. ORIGINAL: I had to put it on HuggingFace for now since this site isn't working today for me: https://huggingface.co/CoachBate/UnderBoner
Thank you very much for your work!
It's listed on the site as T2V but the download is titled as I2V?
I got confused myself there for a second. Underboner is trained on videos for the WAN 2.2 and LTX-2.3 loras and both are i2v. I got it confused with my WAN 2.2 Penis Lora which is t2v
A good add. The closest thing is the Cockshock LoRA. Need more utility LoRA's like this.
Yeah really nice job. Klein is so good at editing that this is perfect for it
❤️🍆
how to get rid of the terrible cursed "music" that seems baked in to the ltx LoRA?
My training data didn't have any music, and I didn't caption them for any music. Lots of things you can do:
you can load my lora with Advanced LTX Lora Loader and set the audio video properties to 0 and just keep the video at 1, OR
easier to just use your prompt to say the kind of background sounds you want or say silence, OR
use workflow that lets you use your own music file background see https://huggingface.co/RuneXX/LTX-2.3-Workflows
Actually, I just checked and this LoRA is image to video and NO audio was set when training since they were pretty silent anyways. That cursed music is LTX and no prompt from you, I assume.
@coachbate More people should know this. That node is frickin' amazing. I have not seen any of the ComfyUI youtubers use or highlight its function yet.
@steezo_jones I've been adding 'no audio' versions in the optional files section for my ltx-2.3 and sulphur2/eros loras where i strip them out programatically.
do they only go to the side? or also up?
you can try prompting it to do so. there were some training images of that. it's more about where it thinks it can put it best so if you have lots of space above and hands down on your thighs that could work. which model are you using?
You can inpaint it to whatever direction or shape you want with the image LoRAs
you can draw exactly the size and direction using the ideogram4 lora
Does this lora you have give an image a boner for the LTX 2.3 one or is it just them touching themselves? I was hoping for a boner lora under briefs or clothes someone plz make it🙏🙏🙏
the image loras give you the boner, the video ones animate it. The LTX-2.3 one is i2v so it requires an image
It seems the Qwen and Flux versions have broken download links :-(
After it give you the bad link error, go back in your browser, and click the download button again and it works. Not sure what's going on with the site now.
Admire your work, man. Any chance to make lora for Qwen 2512?
2512 is my favorite model for my penis lora, but why not just use my lora for Qwen Image Edit 2511 for underboner?
@coachbate I tried, but doesn't work with qwen image 2512 fp8 scaled. 0.3-0.9, nope, trigger word underboner. any advice about prompt?
@JackyLeBroon I meant to say use the qwen edit 2511 model with my lora to do a text to image like you would do in 2512. btw I just did try training it on a lower powered cloud gpu and it made it 2 steps in 2 hours i let it run so gave up on that. i'll eventually get to running it on the 2512 model.
i got it to work with 2512 using the first one i tried, strength 1.0 (maybe that was too low) fp8_e4m3fn_scaled_comfyui_4steps_v1.0
I can post to the showcase. https://civitai.red/posts/29182669