This model was trained on realistic and animated videos on the Wan I2C 720P 14P model. It captures the motion futanari jerking off and ejaculating on her face. I trained on an A6000 using 20 2 second videos (24 FPS, at least 48 frames each). Training 1400 steps took around 24 hours.
Description
FAQ
Comments (13)
What sort of captions do you use with video training? do you provide .txt like with image captioning?
Sure, but don't take me as an expert, because I'm figuring this out as I go. Here's an example of one from this training set:
"black hair, fake breasts, see-through top, bellybutton piercing, tattoo, text on screen"
However, I've since started using full sentences as the Wan video model seems to work best with more descriptive text. Here's an example:
"She is rubbing her vagina with her hand. There is text overlay on screen."
@pikenrover Interesting. Does is also require a trigger word at the start?
I love this Lora and want to use it for "everything," but it has a nasty tendency to turn most videos I toss at it into slow motion. This is most likely because it was trained at 24 fps when WAN itself is trained on 16 fps, causing a speed mismatch. I don't want to be demanding, but I'd really appreciate a version of this Lora trained at 16 fps to avoid the slow motion effect.
I wasn't aware of the fps mismatch, so I appreciate this comment! I will get an updated lora for this at some point, no promises as to when.
@pikenrover I'm no tech wizard when it comes to Wan, so you might want to double-check that what I said is true, but at least I've heard others mention this for other loras.
@RiseKujikawa Wouldnt upsampling fix that?
@PersonOfInterest Upsampling? Are you sure you're using that word correctly? Do you mean speeding up the video? If so, it works, but that just shortens the video, and that's not great for this particular subject matter, which needs a few seconds to serve its purpose.
Liquid tends to flow backwards. It often flows from the mouth to the water gun. What could be the problem?
I cant get this thing to work. im using wan 2.2 and injecting it into both nodes where high an low would go. used trigger and set strength to over 2. Nothing. Pls help.
This lora is not intended for Wan 2.2, and I believe there are existing loras out there that can accomplish this, so I don't have plans to train against Wan 2.2.
@pikenrover Ohhh gotcha, sorry still very new to this, srry for the dumb post. Appreciate it's here!
@maybemalice No worries!
Details
Files
wan-i2v-trans-self-facial-v1.safetensors
Mirrors
wan-i2v-trans-self-facial-v1.safetensors
wan-i2v-trans-self-facial-v1.safetensors
wan-i2v-trans-self-facial-v1.safetensors
125-wan-trans-self-facial.safetensors
wan-i2v-trans-self-facial-v1.safetensors
wan-i2v-trans-self-facial-v1.safetensors
wan-i2v-trans-self-facial-v1.safetensors
trans-self-facial-v1.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.