V1.0
New trigger on this version: "a woman doing a standing full frontbend pose"
But it seems like describing the pose works just as well if not even better,
example "a woman bending forward with her head between her legs standing on a rug"
V0.5
This first version seems buggy, might be bad keywording.
Might work better with longer prompts.
And it seems like WAN has trouble with upside down faces, much like other models
Description
FAQ
Comments (4)
Where do you even find the dataset to make these? There was your pretzel one and now this. Is it trained in only generated images?
Not only generated images, there are yoga images and videos but can take a while to find.
Benkebengt How many clips and images did you use to make those loras? I have a very weird lora I wish to make, but it's so specific that it's hard to find images or vids out there for dataset.
naomixlove The pretzel lora which works much better than the standing one had 13 videos and 6 images. Some images was stills from the videos I used. Mine are trained here on Civitai