Wan 2.2:
This was trained using ai-toolkit on Wan 2.2 I2V 14B using 20 81-frame 16-fps videos.
Wan 2.1:
This model was trained using diffusion-pipe on the Wan I2V 14B 720P checkpoint using ~30 2-second videos from various anal sex positions at 24fps. I used an Nvidia A6000.
I'm aware that it's the general convention to train on the T2V Wan checkpoint, but I get horrendous results (with both T2V and I2V) when I do so with this dataset. My theory is Wan isn't as aware of this type of action as other ones, but I'm not really sure. I just know that I've tried a bunch of times to train this model in T2V to no avail while I've had success training on T2V with other datasets.
Description
Trained on more videos including the best of the best outputs from v0.1. Results should be much more consistent.
FAQ
Comments (13)
Maybe you could write a prompt?
This was actually the prompt used for the video generation: "Moving towards and away from her pelvis"
For the initial image generation, check the metadata on this: https://civitai.com/images/70872685
Hey man, this is a great lora! Are you planning on training taker pov sex/anal?
Perhaps at some point!
@pikenrover it would be great
Great start! It seems difficult to combine with other prompts/loras while maintaining the thrusting movement, though. I hope this can get improved upon over time.
No luck so far getting this to work with wan 2.2. Seems like the viewer just gets carried around like they're in a wheelbarrow with the penis.
time to make 2.2 version xD
Would be possible to have/see the workflow that you've used, with mine I'm getting a bunch of errors:
lora key not loaded: diffusion_model.blocks.8.cross_attn.k_img.lora_A.weight
it's about Wan2.2 I guess, but I'm too noob to understand what I'm doing 😅
A wan 2.2 version would be awesome
Finally got around to making a Wan 2.2 version. Enjoy!
@pikenrover oh awesome can't wait to try it, thank you!
Works pretty well if the original image is already taker pov, if not it's kind of difficult. But a very nice lora, thanks for your effort :)