WAN 2.2
This one is pretty damn good if I do say so myself. Swapped out some of the other penis LoRAs for WAN 2.2 in my tests and this v2 blows them all away. Your results may vary, so look at the showcase prompts for help. This can create videos or images of male anatomy from a prompt, in realistic detail. It can also preserve the genitals if they are in your start frame for image to video workflows. Detailed anatomy prompting helps, see showcase for v1 and v2 for example prompts, or use local LLM Qwen VL or Grok for help with dick anatomy terms.
And it knows cum this time. No other LoRAs were used for showcase videos except for Lightx2v. For the cumshot, describe how much and exact details of where it goes. Beware saying 'massive cumshot' or you'll get a firehose like velocity and volume shooting out (see showcase).
You need both the high and low models in your workflows for video creation. The high noise model is in the Required Components panel. The high noise part gets the shape, and low noise gets the details. You can use the low noise with WAN 2.1 or its variants such as the amazing Bindweave. You can create images also with Magic WAN that are quite good (see showcase).
v2
Built on v1, adding 60 videos and the latest image dataset with the goal to fix the backwards or upside-down anatomy. It still can produce backwards glans, but before it would do it 90% of the time when viewing the underside of the penis, and now completely opposite, failing about 10% of my tests. Prompts can help if you get that. use 'detailed defined glans'. I guess it's hard to get rid of all the backwards dick heads in this world. Trained to a total of 15k steps, 20 hours of additional GPU time for version 2, stopping and adjusting along the way. Loads of work
v1
Trained on the same dataset as versions 1 of my Z-Image Turbo and LTX-2 LoRA, but for WAN 2.2 trained with the T2V model, which can be used to generate penises from your detailed prompt, or keep them in your scene if using i2v workflow. See showcase for prompts and workflows should be embedded in most mp4 files.
Description
Low noise model, you need both high and low for any novices, but have used low noise model in WAN 2.1 workflows and for text to image workflows.
I replaced the 10k step low-noise model originally posted with the 9750-step model since I was getting some double headed penises there.
FAQ
Comments (32)
UniPC | sgm_uniform seems to work with strength 1 on both H/L samples at 5 | No flaccid though, so no transition either ;) Ahh mostly get upside down :( Or ball sacks look like chicken crests.
switched to your lightling, scheduler and sampler even with high samples (10-10) I'm unable to get something that doesnt look like a meat cone.
my videos should have the workflows embedded in the mp4 files unless they are mashups that went through Premiere
i2v or t2v I would use my Flux2 Klein or Z Image Base LoRA to create still images then animate them with WAN 2.2 and my lora so it doesn't disintegrate the dick into nothingness. Klein is nice because you can past the output as your new input and make text based edits to parts you don't like
They have, they are basically the same workflow that I'm using since its derived from the wan template. You use the lora loading and the context windows and the video smoothing and there are traces that suggest you use first frame last frame in other workflows.
I've accomplished to get some erect penis, but none flaccid. I'll correct my comment TYVM for the feedback.
@kcouuock I don't have many flaccid training images by comparison to hard in my dataset, but seems different levels of learning between the various models when it comes to penises anyway. my goal is to have it know the right side up regardless of angle, uncut or cut, soft or hard, and those are still proving difficult to get right consistently, which is why i2v is good. For longer videos like my 2.5-minute bukake trackmeat one, i created a keyframe image and use FLF for the entire thing. When I didn't have a final frame, i had it do i2v and then i captured an end frame I liked, latent upscaled it and edited it and put it in my queue. I have a looping travel prompt workflow that will load lora dynamically for each chunk just off the <lora: > old school tag . Without the keyframes those deteriorate into monsters in 25 seconds.
@kcouuock the wan 2.1 flaccid epoch lora is good along with a flaccid one that is for realistic physics bouncing and swinging and both run on wan 2.2 fine. i usually do 1.5 low and 3 high for the 2.1 lora strength on wan 2.2
@coachbate WIP with your latest comments! I can't add the videos created with this LoRA because are I2V and this lora is set to T2V and the default upload filter doesn't allow to reference to this LoRA.
@kcouuock This lora can be used for t2v or i2v. Wasn't sure which to use on civitai
i found I kept going back to euler/simple after trying all sorts of other combinations. DPM++ SDE/beta and DDIM/beta were good, even LCM/simple or beta. Also check out 'penis arousal' and 'penis reveal' LoRas, one of them is on civarchive.com
@coachbate back to euler but sgm_uniform. cant make it work with SVI. generates "partial" members or none over the cloth.
Your work is amazing I gotta say. I'm new to all of this and been trying to use the WAN2.2 14B T2V workflow in ComfyUI and trying to understand how to use your LoRA to replicate one of the video clips you posted with the exact prompt. I'm unsure if I should be using both high and low noise versions of the LoRA? Would it be possible to share your ComfyUI workflow? The video clips I manage to generate never seem to show a penis at all :(
All of my mp4 files should have the workflow embedded in them so you just drag and drop them on the comfy and you’ll see my workflow the only ones that don’t are the ones I would’ve edited in Premiere, but most of the ones on here are not like that . Make sure you say, my keyword BWC and penis in there too which one are you trying to recreate?
@coachbate Yeah so I'm definitely a noob here :D I didn't even know I could view a post and drag it onto ComfyUI and the whole workflow would appear you're saying? I was just picking a simple one to try and recrate - something like:
https://civitai.com/images/117704903
Like I don't know what exactly to drag onto ComfyUI to get the workflow to appear.
I do see the prompt you used - but I'm not sure if I'm using the right nodes in ComfyUI to use your LoRA I guess is what I mean.
Sorry I'm so clueless! :D
@abrasivetoast Right click on the video from that link you provided and choose 'Save video as ' and put it on your hard drive. The either drag it to the Comfyui canvas, or do File > Open in comfyui and select that mp4 you just saved. btw, the prompt is there on that web page on the right side bottom
thx 😍
But why are they upside down? xD
lay off me. I'm working on it.
Inverted Peens are a problem in Wan. Looking forward to an update @coachbate. <3
@liberty_cipher I'll eventually get to training this with videos too. My focus has been on LTX-2 for video and would need to modify video clips, but new toolkit is out that is supposed to help with that
@coachbate Sounds good. Appreciate the work. Hope the toolkit works for ya.
I keep having lots of problems with that. Even with simply getting a flaccid penis.
@kcouuock I tried every trick in the book and could not get the head of the penis on the right way when it is pointing up and camera is looking directly at it other than about 2 out of 50 times for text to video. You best bet is to use one of my image loras like for Qwen Image 2512 or for Z Image Base, and then use that as your starting frame in an image to video workflow. You get much control that way.
@kcouuock Oh, you mean they are always hard or you get monsters instead of flaccid? My dataset trained with very little flaccid penis but the Qwen Image lora does best at that. Do your keyframes there and use with WAN 2.2 or LTX-2.3
I can't credit you because this lora doesn't appear in the search section when uploading videos, don't know why. Any hints?
can you try now?
@coachbate Now it works, credited ;)
BTW which light loras do you use? I'm unable to make it work with 1030, 1022 or 1217. I only get partial results. works with light WAN2.1 but the others get much better results in motion and SVI (except for camera prompting!).
@kcouuock 1030 high, 1022 low or I use the NSFW FAST move ones that have them built in to the checkpoint. Those will really do things to other guys that my prompting wouldn't do before. I noticed you had a lot of stuff in your prompt. It's only for 5 seconds so better off with travelling prompt unless you can get the newer SVI workflows to work for you. I didn't notice much change but there have been new workflows since I last checked. I still like the looping workflow where i can dynamically load whatever loras i want for each chunk via a prompt and give it a png file with the end frame for that chunk. that gives best results but time consuming
can't generate a penis on wan2.2 i2v. there is only a mishaped genital
see my examples. make sure you have high and low lora model files. You can save my videos as mp4 on here then open them in comfyui. It does help if you generate the penis in an image model where you can get it exactly how you like, e.g. my Flux2. Klein 9B or Z Image Turbo or Qwen Image 2512 or Qwen Image Edit 2511 - I have LoRAs for all of them or use another. My LoRA will then help the dick from falling off or disintegrating, or at least that's the goal.
yea that i mean, it had can generate a penis where is not. T2V or I2V. i used all the options and same settings of the examples, that means it always need a penis in the image or another lora.
@itasky well, the samples that are just images here in my WAN 2.2 model showcase, those are from text only. I can try to help if you share your prompt and starting image
Details
Files
wan22_bwc_coachbate_000009750_low_noise.safetensors
Mirrors
wan22_bwc_coachbate_000009750_low_noise.safetensors
wan22_bwc_coachbate_000009750_low_noise.safetensors
wan22_bwc_coachbate_low_noise.safetensors
wan22_bwc_coachbate_000009750_low_noise.safetensors
wan22_bwc_coachbate_000009750_low_noise.safetensors
wan22_bwc_coachbate_000009750_low_noise.safetensors
wan22_bwc_coachbate_000009750_low_noise.safetensors
wan22_bwc_coachbate_000009750_low_noise.safetensors
wan22_bwc_coachbate_000009750_low_noise.safetensors