I created this mainly to help out a bit with my deepthroat LoRa, so a few of the example videos include that.
It's best at a POV angle, and OK from the side - It can't really do any other angles.
All in all it's a bit limited but figured I'd share it anyway in case anyone gets any use out it.
Description
FAQ
Comments (53)
Bro these example generations are peak lol
This looks great, and I love your examples, just my kind of humour
Hey dude, Trying to get in to generating some loras myself but getting stuck on step 1. Would you know how to run diffusion pipe on google colab? and what the general best settings would be? A colab notebook/tutorial would be great if you could take some time for that.
Sorry, I don't really know much about it and just follow guides myself. This is the one I've followed for doing Loras which uses RunPod:
Love the examples, man xD
I tried to implement one of the workflows from your example images but I'm getting a node missing error on the 'workflowwanparam' node and comfyui can't find it. Any tips on how I can either install the node or whether it's even necessary to include. Thanks :)
Not sure really - When looking at my workflow in ComfyUI I don't have a node by that name, the closest is the node "wanparam". Perhaps something strange is happening with the workflow being saved in the video itself.
Maybe try using my workflow downloaded from here rather than from the video?
https://civitai.com/articles/12380/wan-21-i2v-and-t2v-workflows-12-gb-low-vram-gguf
go back to install missing custom nodes, there is probably a custom node packaged that wasn't properly installed or an incompatible version was installed, tyr changing the version, updating, or deleting and reinstalling, make sure you restart comfyui each time
I had the same issue until I downloaded the workflow OP posted here https://civitai.com/articles/12380/wan-21-i2v-and-t2v-workflows-12-gb-low-vram-gguf When I opened that one it didn't complain about a missing node and the workflow from the videos also loaded correctly after that.
I think it's because the sampler is grouped with an EmptyHunyuanLatentVideo that it bugs out sometimes.
It's really confusing, but the author has create a custom composite 'group node' (perhaps they forgot?). In the linked workflow, you can separate them (right click node and 'convert to nodes') to see what they actually are. Then you can copy them or recreate them in your original workflow.
sometimes you just have an old version of a custom node. Then ComfyUI says it missing, but nothing at the missing nodes tab at the same time. Because you have the node package , just an old version, but comfyUI doesn't tell you that ;). 'Update All' fixed that for me, since all my custom nodes are updated too.
@funscripter627 thanks for that 👍
GOD bless you
I2V possibilities are endless when pairing this LoRa and whatever else you want to happen. I did some testing, it does what you expect.
do you have a specific workflow and the settings/strength for the lora? when i use the lora for I2V it always zooms in and i cannot prevent that
Here you go brother, uploaded this for you. The one I'm using uses SAGEATTENTION and TEACACHE, but if you dont have that installed \ can''t install it I think you can just try setting negative prompts in the workflow you're using now.
The neg that I'm using, which was set by whoever created the baseline workflow I built this off of, is the following:
moving background, moving camera, panning, zooming, tilting, camera move, camera pan, camera zoom, camera tilt, overexposed, bright colors
this is amazing, i am lmao'ing so hard at the videos people are making.
Haven't tried it yet, but yeah, these are the most hilarious samples I have ever seen on here. Too funny. Props to OP
ok i am trying to use this, but it doesnt work, lora does no difference i tried with and without lora and the outputs are exactly the same, i am using 1.3b version for text to video... is that the issue?
It may well be, this was trained on the 14b model, so that could be it?
The clip with the girl drinking coffee could legit be from a national TV ad for Dick. She gets an Oscar (or whatever)
Combined with your deepthroat/blowjob Lora, it works really good, thank you
This is generally great but sometimes give me a slightly "bleaching" of a few frames (i.e., the whole image is washed out), especially used at higher weight. Anyone else getting that? Not sure why that would happen.
yes same happening to me, a flash of white 😭
yes ... same Problem here ... I tried different weights and SLG Settings ... I could't fix it yet ...
I think this Lora works only for 480p ! In 720p you get this strange white blurred Starting-Frames!
Same thing at 480p for i2v. Whenever I used a pre-rendered image or photo, I get a washed out, grain-like effect. It can take upwards of several seconds for the image to settle back to 'normal' but by then most of the prompted action is ruined by the effect. The only workaround is to try and get the video to the desired starting image/action minus the effect, and then use a screenshot as a starting point for what you initially set out to create. Has anyone figured out how to avoid this?
what prompt "trigger word" should i use for cowgirl position? i think do to the angle it not getting detected
What strength do you recommend I use?
I usually use somewhere between 0.6 and 1.00, depending on what other Loras I'm combining it with or not.
What does this lora do?Generate a dick in T2V?What it's gonna do in I2V when there's already a dick.
to keep it a dick.. lol
works for me at least to keep dicks to their proportion/sizes and shapes.
only tried it with some blowjob clips as of now
I've asked this other places, but i'm trying to get a good answer because it's happening with every lora.. why are the penis's constantly off the body or detached from the body? Like i'm getting a penis in the girls hand, or penis on the guys chest.... it constantly happens no matter what I do... wtf?
did you ever figure this out lol. it's so annoying
Its completly repaiting my image when i use this
I'm using images with existing penises and I want to use this lora to keep them at basically their proportion and characteristics so other stuff can use them.
But I'm finding more often than not the lora completely changes the penis, sometimes in warped ways.
Any recommended strength and prompt language to make it more likely that the dick stays in the ballpark?
This v1 Version seems only be 480p - Right? If yes, Please do a 14B 720p Version! Thank you!
Very useful actually. Thank you
is it a lora or can i put it in stable diffusion?
If you use one of the sample videos to load the workflow in comfyui and get an error for 'workflowwanparam', it seems to be a group node you can replace with EmptyHunyuanLatentVideo and KSampler.
Is this the cursed lora that fucks up the first frame by making it grey and gives dicks to women?
Is it? Samples don't show that, are they rigged?
we need a 720 p version
IMPORTANT: INFORM ALL WAN USERS! I DISCOVERED HOW TO USE LORAS 'NSFW/SFW' IN WAN 2.2 T2V, GETTING THE FULL POTENTIAL OF "HIGH NOISE" AND "LOW NOISE." THE RESULTS WERE IMPRESSIVE!! I ran it on an RTX 3060 6GB, 32GB RAM, 480x480, length: 41, generation time: 5 minutes to generate each video. The quality is fantastic! For those of you with more powerful computers, the results will be even more incredible. I used the models: WAN 2.2 - Q4_1.gguf FOLLOW THE STEPS BELOW: 1: Don't use acceleration/enhancement LoRas (such as Lightx2v or other) on "high noise"; use only on "low noise" with strength 1. (I'm using this version of Lightx2v LoRa): lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank64_bf16 2: Use 'NSFW/SFW' LoRas on both "high noise" and "low noise" models. On "high noise," use strength 2, and on "low noise," use half the strength, with strength 1. You can vary the strengths as you like, but always keep the "low noise" strength at half the strength of the "high noise." 3: Use these settings in KSamplers: "high noise": steps: 9 / cfg: 3.5 / euler/simple / start_at_step: 0 / end_at_step: 4 "low noise": steps: 9 / cfg: 1.0 / euler/simple / start_at_step: 4 / end_at_step: 10000 Note: To improve speed, I set the Windows system to "best performance," disable unnecessary background programs and applications, and lower the desktop resolution. This significantly speeds up video generation. I hope this helps.
This comment is a diamond in the rough - dropping lightning Lora from 2.2 hi noise pass and doubling action Lora strength has changed the entire game for my workflows - thank you for sharing.
@k1337install2300 you can also do a quad-ksampler setup. So I do 4 steps high (no lightx lora) then 2 steps high (with lightx lora), then I do 4 steps low (no lightx lora) and a final 2 steps low (with lightx lora)
I'm only now going to try incorporating some of these wan2.1 loras into this 4-Ksampler setup.
At first I will only add this lora to high noise, as that's important for movement. But I'm thinking it will need it on all four Ksamplers.
I will post a workflow to my profile when I get it working!
Holy shit, you've just dropped my gen time to a third of what it was and the quality is the same! You're a hero!
This setup will change the faces of your subjects. The effects of the loras are stronger and that may work for clips with random people but if you work with someone you can recognize, you will probably end up with a different person.
You have been warned. Good luck.
PS: for speed I use Torsten workflow. Try it.
Sounding Lora when?
I'm new to WAN and I can't figure this out. I put the LORA in at 1 model and 1 clip strength and it didn't seem to do anything. the dick just looks like an arm with the hand amputated...
What's the right way to use this?
wan2.2 14b version when
no 2.2 version yet :(
Details
Files
dicks_epoch_100.safetensors
Mirrors
dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
123_dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
wan_dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
dicks_epoch_100.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.