WAN 2.2 I2V - POV Cowgirl - CivArchive (CivitAI Archive)

WAN 2.2 I2V - POV Cowgirl - LOW v1.0

NSFW

UPDATE 9/1 v1

Updated training settings should give even better movement.

---
UPDATE 8/19 v0.2

I've updated the dataset & training settings so that movement is more preserved. It works much better even without the general NSFW lora, give it a try and let me know how it goes for you.
----

My attempt at POV cowgirl for WAN 2.2. It works pretty well (doubly so if you also use the general NSFW lora), though the girl's hip motion can be less aggressive than I'd like.

This is only my second ever video lora and I'm training locally on a 3090 so it's slow progress. I'm still tweaking the training settings and the training data to hopefully show more hip movement. Give it a try and let me know your thoughts!

Description

FAQ

Comments (19)

xjtomb548Sep 1, 2025· 1 reaction

CivitAI

The 1.0v generated is not as good as 0.2, and it can't perform any action at all.

TwoMoreLurker

Author

Sep 1, 2025· 1 reaction

Yeah woah you're right, I must have gotten the files mixed up. Give me a bit and I'll reupload.

TwoMoreLurker

Author

Sep 1, 2025· 1 reaction

@xjtomb548 Fixed the files now, can you check if they're working for you now?

xjtomb548Sep 1, 2025· 1 reaction

@TwoMoreLurker Yes, I can check them now.

xjtomb548Sep 1, 2025· 1 reaction

@TwoMoreLurker I've tested it, and it's much better now.

TwoMoreLurker

Author

Sep 1, 2025· 1 reaction

@xjtomb548 Thank you for pointing out the error & testing the fix!!!

likaikou3588113Sep 1, 2025

CivitAI

no one try it?

duzhongzhuSep 2, 2025

CivitAI

let me try try

illumizoldyck808524Sep 10, 2025

if it's working for you then can you pls share the workflow.

koopcco768Sep 2, 2025

CivitAI

No prompt words? If there was no penis in the beginning of the scene, how would he behave?

TwoMoreLurker

Author

Sep 3, 2025

The dataset is all captioned with natural language so there's no "trigger word", just write a descriptive prompt and it will work.

I don't know what'll happen if she's not already on top at the beginning of the scene, it's outside the scope of what I trained this lora for. Maybe give it a try and see what happens?

hereandnow931Sep 3, 2025· 10 reactions

CivitAI

What is the difference between high and low?

Light6969Sep 3, 2025· 7 reactions

With Wan 2.2 models there are two models, high and low, you usually want to use a lora for each one. High tends to focus more on the motion quality, then Low finishes off the video by adding the details.

fenasikerimSep 4, 2025· 1 reaction

@Light2020462 Isn't it exactly the opposite?

Light6969Sep 4, 2025· 3 reactions

@fenasikerim I don't think so.

NeuralNudesOct 1, 2025

CivitAI

why is it often missing the tip of the dick when it comes out?

robycOct 21, 2025

CivitAI

Can this handle insertion too? I have a pic I would really like to use this for but I need that movement before

TwoMoreLurker

Author

Oct 21, 2025· 1 reaction

Probably not that well? I haven't tried. You might have luck combining it with the POV insertion lora.

ClocksmithOct 26, 2025· 2 reactions

This works well from my testing. I used the HIGH A and LOW models with good results:
https://civitai.com/models/2048863/wan-pov-cowgirl-insertion

LORA

Wan Video 2.2 I2V-A14B

by TwoMoreLurker

Download (Beta) View on CivitAI

Details

Downloads

12,659

Platform

CivitAI

Platform Status

Available

Created

9/1/2025

Updated

5/15/2026

Deleted

Files

WAN-2.2-I2V-POV-Cowgirl-LOW-v1.0-fixed.safetensors

Size:

585.10 MB

SHA256:

5f925dd66203c75256bf85c989569c8445c8299066faedf1c4cad5e407faa044