V1 (WAN)
WAN version is here! Somehow, someway I was able to get this running on 2 epochs at only 200 steps.
I use Hearmeman's workflows for both t2v/i2v. Highly recommend. I do use Tea Cache but haven't tried Sage Attention so it may or may not work. You can also drag the gallery sample videos into the workflow for the exact prompt and settings I used.
Trained on Wan 2.1 T2V 14B. I2V isn't as good. Might have to experiment.
When prompting add "woman lifting up her shirt exposing her breasts" and "**********". You can change the article of the top and to a bra instead of breasts if you'd like.
It's not perfect but much much better and realibe then the Hunyuan version.