Wan 2.2:
Download both the highnoise and lownoise loras, and plug them both into your workflow. The "training images" download contains the example I used for the previews.
T2V notes: I did not use the self forcing lora for the highnoise part of the workflow, because I find that it limits facial variety too much. In my testing, this happened with or without the missionary lora plugged in, so if you're having trouble with facial variety, you may have to disable the speed lora for the highnoise step.
I2V notes: This version should handle cases where the penis starts outside the vagina. It can handle cases where there's no man at all in the starting image, however, don't expect the penis to look very good in these cases. Wan still doesn't have a good idea of what they're supposed to look like. As with the T2V version, I found that using the speed loras for the highnoise step causes the results to be less good. For 5 seconds of video, the lightning lora seemed to cause the videos to be in slow motion when using it on the highnoise step. Maybe there are good settings that I just haven't found yet.
Wan2.1:
I2V version: While the T2V model can work really well as I2V, it seems to be picky about resolutions and aspect ratios and things like that. The specifically trained I2V model just seems to work better in more situations. So, it's becoming clear to me that providing an I2V model along with the T2V model is worth it. It uses the same training data as the T2V one, minus the static images.
Version 1.1 Updates: Trained at a higher resolution in an attempt to get higher quality outputs. I also added more training data so there's more movement and bouncing breasts.
Important parts of the prompt:
with her legs spread having sex with a man
...
A man is thrusting his penis back and forth inside her vagina at the bottom of the screen{Movement is fast with bouncing breasts|Movement is slow}Her breasts are {small|medium sized|large}Sometimes it may help to put the things you don't want in the negatives. For example, if you want small breasts, putting large breasts in the negative could help.
Description
FAQ
Comments (61)
i2v version for i2v does seem to work better than the t2v model. meaning that while t2v loras do work for i2v it is still better to just train both. interesting. would be nice if other lora makers could verify this.
Can I use this with my 4060 8G graphics card? Or is there a recommended configuration?
I've personally had the best results using this workflow, in ComfyUI, so far - using the specified 480p GGUF model - https://civitai.com/models/1309369?modelVersionId=1613089
Yeah it's not going to be about the workflow so much as using a model quantized to a size that will fit on your card: https://huggingface.co/city96/Wan2.1-T2V-14B-gguf
i would try 3_K_M/S or the 4_K_M/S if I were you, but if you have say 40GB of RAM you technically run the entire 30gb model. It would just take a really long time.
Someone please nominate this guy for a nobel prize! (new category). Thank for this amazing lora! it's be amazing if you'll plan to train some other common position too like fullnelon / facefront doggy. But take ur time, we are more than grateful already! keep up the great works!
For some reason the face ends up being very low quality for me.
If you're generating the starting image, make sure the face is good there. I noticed that if the eyes are a little weird, it picks up on that and keeps it that way. I had to bump up the face detailing on the starting image and that seemed to result in much better faces.
use ROOP to fix the face, it could be anything causing it, if the face deviates from original image it could be you're generating at low res
I've found with Wan I2V that making sure the aspect ratio / resolution of the input image matches something that the base model or lora were trained on makes a huge difference in face quality (and quality in general.)
Great work, can you try a hand job WAN please? preferably i2v
I only do I2V and I've been using both this and the I2V v1.1 version and I find that I consistently get better results with this T2V version.
Hmm, interesting. What resolution and aspect ratio are you using?
@dtwr434 480p at 9:16
@xXDMTXx Odd. Well as long as something is working well, that's good I guess. When you say you get better results, do you mean the overall motion, or the quality of the image?
@dtwr434 the thrusting motion is more consistent. With the i2v version, the motion falters with girls just hovering over the penis
I have had the same experience. The motion is more energetic and consistent with T2V version. (Which I assume is the OG 1.1 version?) The I2V 1.1 is not bad and it seems less prone to artifacts, but the overall motion is not as good/satisfying.
@VorzKorz Alright. When doing testing on certain resolutions/aspect ratios, it was hard to tell if there was much of a difference. I think the I2V version might be better for more non standard resolutions, so your mileage may vary.
@dtwr434 Thanks for the reply! I see. I am using 512x768 (portrait) aspect images as the starting point. Although they are higer-res than that, because I read in multiple places that having a very high-res starter image is supposed to help.
for some reason my outputs keep generating with tthe male fucking their belly button or just above wher their vagina would be :(
well, i have the same problem. anyone know how to fix it?
Alguuen ayudenme hacer un video no encuentro las conexiones
So the range of motions and expressions on the girl are very good, but I've found a bug. An anime-ish girl with big eyes with a bright colour would close them in the first few frames and when she reopened them they were no longer big, but just normal dark brown eyes.
So tell it about how it is anime. I use, "The video keeps the same Anime style with big eyes and small mouths."
Gonna have to say hell yes man great job
The body motion is fantastic, but it changes the faces on my images so the likeness is lost. Would love a version that is trained with no faces so it is face preserving. The only issue I see. Hoping you can work on a fix for it!
I put, "changing facial features, becoming a different person," in my negs and "She keeps the same beautiful facial features with eye makeup and green eyes and a slightly big nose and thin lips" or whatever in the prompt.
trying to make that work on my pc, but seems like it cant be used with my model 1.3B_bf16... is that right? They dont fit together?
That's right. This lora is for the 14b model.
This LoRA is a dream come true! Generating a video that results in a decent loop takes some finessing, but I am satisfied with the results!
Check out the loop anything workflow, it's been great.
A good number of times I run this lora, the man ends up with a miniscule penis, literally as narrow as a pencil. What's a good way to solve this?
Hmm, honestly I have no idea. I haven't encountered that myself.
penis LORA 能有效解决
@xuelang_8899802 got an url? that name does not work on civitai.
I was also troubled by the problem of my penis turning into liquid for a long time. It didn't work with the penis LORA. Finally, after I cancelled the LORA one by one and did a step-by-step test, canceling the use of this LORA restored 90% of the problem. I don't know why the problem of a small penis would occur when using multiple LORA conflicts.
@s85157225117 wow into liquid? interesting, i still have many issues with penises in image gemeration. sometimes they look like creepy tentacles, sometimes they are like thin pencils, sometimes they look just deformed, and very often they will not be attached to a male body and just floating, or attached in the wrong spot. would love to have a discussion with people who know how to actively solve these things.
some asshole stole your lora and videos and reposted it
Hope you can update this fantastic LoRA to use 'Wan Video 14B i2v 720p' base model since the latest generator changes.
Hi, it seems nothing happens when i use the trigger words.... girl is doing nothing (lloking around) no movement. Any tips??? Thanks!
Ah... the guy's position must be at the bottom.... no exceptions, i will try some more.
@phntalsma544 Thanks mate... its working
Does anyone know a good 720p missionary lora? This one has a hard time as it's 480p and doesn't fully penetrate
I'm seeking 720p right now...
mickyIammicky I saw this one last week, haven't tried though. Mind trying it and reporting back?
https://civitai.com/models/1745090/wan-dr34m15h-multi-perspective-missionary-pov
ItsThatTimeAgain You just found what I need. I try it today. I'll tell you then.
ItsThatTimeAgain Above 720p, it just doesn't work—the whole video gets completely corrupted.
mickyIammicky Hmm the minimal size need to be 720pixel when generating the raw output with the 720p model. Same thing for 480p model, it need that one of the values be maximum 480p and the other doesn't matter. For example if you want a final output of 1080p (1920x1080), you need to generate at 1280x720 and then upscale 1.5 times.
Mind trying it at 1280x720 or vice versa 720x1280 portrait to see if it works?
ItsThatTimeAgain I appreciate your guidance—you’ve cleared up the doubts I’ve been having lately. I really feel I can learn a lot from you. Let me try again and see if it works.
ItsThatTimeAgain ItsThatTimeAgain By the way, I don't wanna do 1280x720/720×1280, no patience for the waiting LOL. Cuz it will take so much time to gen. I do it at 720×720, I guess it works too right?
mickyIammicky Yeah that can work, but honestly, if you just want something fast you can use the 480p version and upscale it. I use 720p as illustrated characters lose lots of quality and create artifacts on 480p. I found 480p doesn't have this issue when creating realistic characters and backgrounds.
ItsThatTimeAgain Hey, I tried the LoRA - it works fine at 720×720. Well I need more than just speed; I want over 90% character consistency and movement control, plus no pixelation on character's eyes.
mickyIammicky Awesome thanks for testing it, I'll add it to my position lora list. Yeah 720p is indeed what you wish for HD results
lora works great, I really like the movements I get, but there is an artifact at the bottom in the form of a dark band (like a table surface or something)
I use with i2v, 720p
check your prompt for miss-spelled words or words that the ai doesn't understand
Need node Int
Hi, love your lora. I am new to AI video generate, may I ask how can I generate a good result if the start image don't have man's body or dick?
That probably won't work well, as all of the training data started with about what you see in the samples. If I ever make a new version of this, I'd probably try to include something to handle that, but I'm not sure when I'd get around to it.
dtwr434 Got it, anyway thank you for this awesome lora!
this is a little rhetorical but you could use a mask to add the penis, right? Or at the least I2I the image, then use that?
Are you considering updating this fabulous Lora to Wan 2.2?
Perhaps eventually, but it may take a while to get to since it looks like it requires training it twice for both I2V and T2V. In the meantime, it still seems to work somewhat decently as-is with 2.2.
Details
Files
wan_pov_missionary_i2v_v1.1.safetensors
Mirrors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
【wan】躺插pov_missionary_i2v_v1.1.safetensors
wan2.1-i2v-s-miss-pov.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan21-action-missionary-pov-i2v-v2.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan21-action-missionary-pov-i2v-v2.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
wan_pov_missionary_i2v_v1.1.safetensors
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.