Wan Front View Doggy Style (i2v) - CivArchive (CivitAI Archive)

Wan Front View Doggy Style (i2v) - v1.1

NSFW

UPDATE: v1.1

v1.1 has significantly better thrusting motion from the guy, overall I think it is much better than v1.0

All the showcase videos were generated image2video with 10 steps with the prompt:

"Pl0wView. Video of a woman being fucked doggy style by a man from behind. Her face expressing intense pleasure and moaning. The man aggressively thrusts deep and hard into her rhythmically causing the woman's ass and breasts to bounce and jiggle in rhythm with his deep and hard thrusts."

v1.1 was trained on the 480p i2v 14B model with diffusion-pipe on 22 videos for 30 hours on two 4090s.

-v1.0 - WORK IN PROGRESS

This model is for doggy style action with the woman facing the camera. In my testing I found the faces generally turn out pretty good, however control of the man's movement needs work. I will be working on improving the data and coming out with v1.1 sometime. This initial training run took 22 hours on two 4090s.

Trigger word is: Pl0wView

Try it out and let me know what you think. Feedback on what works well and what sucks is appreciated. I generated about 50 videos in testing but it takes a long time so I haven't tried out everything.

Description

v1.1 was trained with diffusion-pipe on 22 videos for 30 hours on two 4090s.

FAQ

Comments (39)

crombobularMar 31, 2025

CivitAI

trained on i2v or t2v?

lazerblazer

Author

Mar 31, 2025

It was trained using the i2v model. I don't know if that means t2v won't work. If you (or anyone else) try it with the t2v model let us know if it works or not, I'd appreciate the feedback!

crombobularMar 31, 2025

@lazerblazer yeah usually that means it won't work as good. the recommended training is always to use the t2v model because that works on both the i2v and t2v whereas a lora trained with i2v only works on i2v

lazerblazer

Author

Mar 31, 2025· 5 reactions

@crombobular Ah, I didn't know that! I am already getting a data set together for my next lora so I will test out training on the t2v model.

logenninefingers888Apr 1, 2025

@lazerblazer while T2V loras work for I2V I've heard I2V loras are still better (for I2V). I2V loras don't work for T2V at all though I think (base Wan just is just too much of a prude).

crazybabyApr 1, 2025· 1 reaction

CivitAI

nice！

SLAPY7Apr 1, 2025· 4 reactions

CivitAI

Can you please do a t2v version :(

lazerblazer

Author

Apr 1, 2025

No, this took a long time to put together and I think it turned out quite good for i2v, which is what I personally use. It is what it is. Other users have reported that my other loras work for t2v, so it may be worth trying, however I don't use t2v so it is untested by me.

CyberAImaniaApr 1, 2025· 2 reactions

CivitAI

Very not natural. Mostly small head , big body. Like mutants

logenninefingers888Apr 1, 2025· 6 reactions

thatsmyfetish.gif

lazerblazer

Author

Apr 1, 2025· 3 reactions

That's just my source images used for the image2video generation being weird. My image workflow has a lot of trouble getting good front view doggy-style poses with normal proportions (it's a skill issue). All the training for the lora was done on real life videos so the proportions it's trained on are natural human proportions. If your source image has normal proportions, the model won't change them.

effyaiaiApr 16, 2025

CivitAI

can anyone help me make video on s d

argoticMay 11, 2025

CivitAI

how many frame per second do you advise to have a great motion ?

lazerblazer

Author

May 12, 2025

WAN’s default fps is 16, so I do first run at 16 fps then I interpolate up to 32 fps. Not sure if that’s the best way or not.

FutopiaMay 14, 2025

CivitAI

Great LoRA! Thanks. I only wonder what is the proper prompt or what TextToImage LoRA does the standing bent over pose? The one with the bum up and hand and feet on the ground.

superwai777974Jun 7, 2025· 3 reactions

CivitAI

Please change your base model of this lora to wan 2.1 720p & 480p so that we can use this lora in new wan generator thank

lazerblazer

Author

Jul 9, 2025

Just now seeing this, the model has been updated now so hopefully it works on the generator.

GinBuddy57746Jul 8, 2025

CivitAI

It won't load when using wan2.1-i2v-14b-480p-Q8_0.gguf for some reason. Does it only work with certain models? So far, this is the only LORA that throws an error.

luffyd1900566Jul 20, 2025

CivitAI

Tested it, it was good! although it has a problem with the eyes, it seems like it distorts it when moving, even with CFG10 and 80 STEPS. Looking forward to see it fix in the next update! TY for this incredible work!

1661204908Jul 25, 2025

CivitAI

Why do female characters also generate a penis when I am i2v?

1191300911Jul 30, 2025

CivitAI

I put the video you showed directly into comfyui, and the result is a "Upscale Factor" node missing from the workflow, and its output port is "INT". I didn't find the missing plugin in the Manager, how do I get it or a node that has the same function as it?

lazerblazer

Author

Jul 31, 2025

In that workflow the node is just passing a 2, so any node that you can get out output an INT value of 2 will work. That specific one is from ComfyLiterals. I can see it on Comfy Manger, not sure why its not finding it automatically.

1191300911Aug 2, 2025· 1 reaction

lazerblazer Thank you so much! This helped me a lot as a beginner!

FlowMowAug 4, 2025

CivitAI

I am getting a

mat1 and mat2 shapes cannot be multiplied (77x768 and 4096x5120)

error when using the wan2.1. Which model was this used with to train?

lazerblazer

Author

Aug 5, 2025

I have not gotten that error before, hopefully someone else can help. It was trained on Wan2.1 480p i2v 14B model

p34313Aug 30, 2025

way too big resolution and try with image to video wan model

LighterAIDec 18, 2025

change load vae input to wan 2.1 tensor

nikecoy884925Aug 10, 2025

CivitAI

what is the seed?

OdbastardAug 17, 2025

CivitAI

How do I test

alternativaonlin3621Nov 13, 2025

CivitAI

Does it needs to put the lora on both high and low?

LighterAIDec 18, 2025

CivitAI

works on wan 2.2?

lazerblazer

Author

Dec 18, 2025· 1 reaction

It works, but its not very good on 2.2.

VelouraMar 17, 2026

if you use a mixed diffusion model like smoothmix, it guides the motion well enough

jdm55555Jan 17, 2026· 1 reaction

CivitAI

Works great with Svi/smoothmixv1!

teeay25225Apr 23, 2026

CivitAI

It wont work and i used 480p + your example prompt. I tested it with checkpoint boundbitehighspeed (thats with the speedlora) and without speed lora: boundbitetruevision. I tested both with 10 steps. The strength on the lora high+low i tested 0.8 and 1. What am i doing wrong? I'm kinda new to this but i wanna learn.

teeay25225Apr 23, 2026

It doesnt work, i tested also cfg 1 and 2.8.. no matter what. i dont get it.

teeay25225Apr 23, 2026

You use the lora only on high if you only have 1 file right?

lazerblazer

Author

Apr 24, 2026

This lora was trained on the base I2V Wan2.1 model. I do not know if it works on other checkpoints, I can't help you there.

lazerblazer

Author

Apr 24, 2026

@teeay25225 This lora is for Wan2.1, not 2.2 where there is separate high/low models. I don't know it works for 2.2, probably not very well if at all.

LORA

Wan Video 14B i2v 480p

by lazerblazer

Download (Beta) View on CivitAI

Details

Downloads

21,607

Platform

CivitAI

Platform Status

Available

Created

3/31/2025

Updated

5/13/2026

Deleted

Trigger Words:

Pl0wView

Files

front_doggy_plow_v1_1_wan.safetensors

Size:

513.87 MB

SHA256:

be3921fbba30f36902898a5ffb209ee3aadfa7de8fae85dcf18edfc6af197ebb