Wan 2.2 update. The preview images were all created using the lightning loras. The workflow can be downloaded in the "training images" download.
The main part of the prompt is:
A woman is straddling a man and having sex with him. She bounces up and down on top of the manDepending on the angle of the camera, you can either directly see the genitals or not. You can add this to your prompt to try to get them to show up:
You can see the woman bouncing up and down on the man's erect penis.You can prompt for different angles:
{The view is from the front and at a 45 degree angle|The view is from the side|The view is from behind at a 45 degree angle|The view is POV}By default, the woman's knees will probably be pointed down, so if you want more of a squatting position with her knees up, you can prompt for that this way:
She is squatting with her knees raisedDescription
Retrained with a higher quality data set and at higher resolution
FAQ
Comments (47)
How are these types of videos made? With stable diffusion?
The WAN video model: https://comfyanonymous.github.io/ComfyUI_examples/wan/
One of the simplest ways must be WanGP 4, it still eats 32gb of system ram though.
I'm not sure I experienced this in versions 1-1.2, but in version 1.3, some of the poses/views are picky about resolution. When asking for a sideview at 45 degrees, 75% of the video is a large black bar when using a tall perspective video. I don't think I had that behaviour on the other two versions. When using the right perspective+size ratio, it works well
Huh, that's strange. Most of the training videos were identical between the versions, just at higher resolution than before. I cut out a few of the lower quality ones. I wonder what would be causing that... I can't say I've noticed that myself, but maybe I need to do more testing. What was the exact resolution you were trying to do?
Incidentally, I'm finding that using a tall perspective is giving me better overall success rate with anatomy in general.
@dtwr434 The tall one was: 480x640
To be fair I have a few additions to the prompt, also using that detailz lora, that may have forced it to pick a particular thing. They aren't that much though, normally they do almost nothing to the examples. After a couple tests it seems maybe the detailz lora was maybe the problem. Lol. Gotta stop using this one I think, at least not first.
The quality is improved so much with 1.3.
even in 256x384 res the result still sharp and detailed not blurred, good job.
Thanks for all those very good trainings. And thanks a LOT for taking the extra time to copy-paste your prompts here and write down a detailed description. Thumbs up.
The latest version has been a blast to use - great job and 2 thumbs up! :)
Keep getting this error on new version
"It seems that models and clips are mixed and interconnected between SDXL Base, SDXL Refiner, SD1.x, and SD2.x. Please verify."
Really consistent across angles vs position. Good job
Im having a hard time with clothing removal on subjects. Im new so bear with me please. If im trying to set up a generation like these but the subject has a shirt on or underwear, how does this LoRa get the panties to move aside or the top to come off? When I prompt for that the generated product looks like it has no idea what to do. Is there a process to go through to remove clothing first?
Removing clothes is a different concept that this lora probably won't be much help with. You'll have to use a starting image that already has the clothes the way you want, or use a different lora specializing in removing clothes first.
with image2video your starting image has to be pretty close to the desired video
Try using the WAN General NSFW lora. It tends to be pretty good at clothing removal; I typically have better luck with higher weights than 1. If still having a hard time, I find sometimes adding the NSFW/Female Genitals Helper lora at low weight can produce nipples and other genitals, just make sure you're describing the desired appearance i.e. "small pale nipples". Good luck!
Please a Hunyuan for framepack T.Tpleasse
How do I get this to work? It says (Incompatible with current model!)
How can I avoid making exaggerated expressions in videos that cause the characters' faces to look distorted?
she unchanging expression that remains naturally consistent throughout the scene.
I try lora strength 0.6 better
Could you upload a few of these models to tenorart? Saw you did the cumshot one? Tenor is missing some sex positions and yours are the absolute best.
Yeah, all of my Wan ones are available: https://tensor.art/u/827267506719714201/models.
Is this just T2V or can it also be used for I2V?
This is mainly used for t2v, and the effect of generating i2v is not ideal. A separate model needs to be trained for i2v.
I used this with i2v previously with some success but it seems like it's not even available anymore.
I changed it to say it's I2V so it can be used on the generator, but it was trained a T2V. It should work for everything if you download it yourself though.
@dtwr434 Thank you!! I am still learning and haven't successfully uploaded a model yet. I'm sure it's easier than I am making it.
At one point I was able to use this for online generation with "BustySigrid - Wan" and "Bouncing Boobs - Wan 14B" to make a popular video that has the most reactions of any I've made, but now for some reason I can't use those 3 LORAs together, what happened?
I think all of the different WAN categories CivitAI has been making has basically ruined the online generation, as I suspect you can't combine things in different categories anymore.
dtwr434 Yeah, it's really goofy and finicky about its auto-detection for resources. It thinks resources aren't compatible even when they are. I've been frustrated by that when uploading stuff I generate locally and trying to list the resources used
can someone please help me with how to setup this.
I use comfyUI. Any help is really appreciated.
I'm using img2vid to generate 0.7 to 1.0, and although it does work, it's too slow. Maybe the seed is off, but even after generating about 40 images, there's no change.
Any plan to create for wan 2.2??
Maybe eventually, but it looks like you have to train it twice each for T2V and I2V, so it seems like kind of a pain. For the moment, the existing loras seem to work decently well already, at least for T2V. I haven't tried I2V.
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.
Oh, thanks! The finished video is way beyond what I expected!
Which part of the workflow do you use lora?
Not loadlora, right?
Input photo is just a bust, how would i make it turn into a cowgirl position. Does anyone have prompt like that, thx.
WAN 2.2 Please!
🥺Does anyone know how to solve the problem of the penis appearing directly in the vagina and being as soft as liquid, and finally merging with the penis below?
#brandnewsentence
can someone provide a NSFW workflow? every video i try to make, the penis seems to "shake" rather then go in and out. both for cowgirl and doggy style videos. have yet to see penetration after 3 days of trying to make video.
My solution is to add some violent jumping to make the picture have a larger displacement so that the penis can be inserted more easily.
Having trouble with faces in this one. They tend to distort quite heavily.
I usually try to lower the weight of the Lora as much as I can in that case... just keep lowering slowly until things no longer work and you should have the sweet spot of still working, but affecting the face the least.
I'm using 2.2 I2V, but I still put this LORA in my hard drive, because it still works well, unfortunately I have to set the intensity under high noise reduction to 3 to better control the angle, otherwise no matter how I modify the prompt, the angle is always wrong or even deformed, compared to many other cow girls' LORA, there are 2.2 and 2.1, this is the best motion control, I hope you can update the 2.2 version as soon as possible
Details
Files
wan_cowgirl_v1.3.safetensors
Mirrors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
102_wan_cowgirl.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
wan_cowgirl_v1.3.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.