Update 1/19:

Version 2.0 was retrained on a larger video dataset.

Version 3.0 are helper loras trained on image only dataset. These loras to be used with the Version 2.0 loras. These loras work by themselves but you might not get the motion/ penis shape you're looking for.

Works on both T2V and I2V.

Sample workflow is provided in the attached videos. (Drag and drop the video into your comfyui tab)

Training details are available here: https://civarchive.com/articles/19881

Long prompts are not required. You can keep it simple with a description of the penis you want to see and it should work.