This lora was trained on clips of the face-down ass-up position, mostly shot from the front and sides. It is focused on the woman's face, breast, and hand movements.
There are 2-3 clips of penetration in the dataset so combined with the general NSFW lora it can probably do POV/behind angle, but by itself it's not good at penetration.
Description
FAQ
Comments (20)
effect is very good
Hi, any chance you could do a Text 2 Video version?
Pretty please?
yes t2v pls
@MMOFan Okay but only because you said pretty please. Give me like 2-5 days, it takes forever to train these things.
@MMOFan I've been working on a T2V version for the past day and a half and finally tested it tonight, and it turned out terrible. I must have messed up the training settings or something because yeesh it was generating some really weird shit. I'll give it another go later this week/early next wee.
@TwoMoreLurker thanks again for making original poses, this one has so much potential.
could you please share a workflow for it?
I have shared my current WAN 2.2 workflow here: https://civitai.com/articles/18162
If you're just starting out I'd recommend the Lightning version as it generates videos significantly faster.
Not sure what i`m missing , have only the face in the video no matter what prompt i use . Much appreciated any guidance to make this work .
This is an I2V lora. Are you starting with an image? It's not meant to work with just a text prompt.
yes , updated with multiple types of images , i can only get that particular model in the image to move , nothing else added ( the man , nudity ) . Tried with different strenghts , cfg ,steps , with or without lightnings , Sry , not really an expert on this , must be doing something wrong i guess :)
@cristianlyon67811 What do you mean "that particular model" in the image to move? I'd suggest using a workflow that is embedded in one of the videos in the examples. You don't want to adjust the CFG for the high noise over 3.5 at any point and keeping the 2nd pass and low noise set to a CFG of 1 is recommended.
Usually the issue comes down to not using the near exact prompt that the lora is trained on. This is due to some loras being overtrained on particular phrases. That's why if a lora includes a trigger word, it is HIGHLY recommended to use it someone in the prompt. EVEN IF IT DOESN'T DESCRIBE WHAT YOU WANT. Just use the trigger to "activate" the lora and this usually fixes the issue.
Like the NSFW movement loras, they work much better if you include their trigger word. Without it then it won't always work properly but with it then things go better.
@cristianlyon67811 bro did you manage to do it?
@lighterai yes , a much better starting image and it works like a charm
@cristianlyon67811 i guess i need a head down ass up picture to start?
@cristianlyon67811 would you share your workflow pls
whats the prompt? cant get them to do anything
The one of the clips in the Gallery below: "A sexy woman is having sex with a man in the face-down ass-up position. She is having sex with a man in the top-down bottom-up position."
What am i missing bro. I never get the face down ass up result. i get irrelevant results. i am adding the screenshot of my workflow. this work flow works fine on other loras but yours. https://ibb.co/LzdTs7sF
i dont have a checkpoint in my workflow. should i use one?
Details
Files
WAN-2.2-I2V-FaceDownAssUp-HIGH-v1.safetensors
Mirrors
WAN-2.2-I2V-FaceDownAssUp-HIGH-v1.safetensors
WAN-2.2-I2V-FaceDownAssUp-HIGH-v1.safetensors
fdau_high.safetensors
127_FaceDownAssUp-HIGH.safetensors
WAN-2.2-I2V-FaceDownAssUp-HIGH-v1.safetensors
WAN-2.2-I2V-FaceDownAssUp-HIGH-v1.safetensors
WAN-2.2-I2V-FaceDownAssUp-HIGH-v1.safetensors
WAN-2.2-I2V-FaceDownAssUp-HIGH-v1.safetensors
WAN-2.2-I2V-FaceDownAssUp-HIGH-v1.safetensors
WAN-2.2-I2V-FaceDownAssUp-HIGH-v1.safetensors
WAN-2.2-I2V-FaceDownAssUp-HIGH-v1.safetensors
WAN-2.2-I2V-FaceDownAssUp-HIGH-v1.safetensors
WAN-2.2-I2V-FaceDownAssUp-HIGH-v1.safetensors