V2 -------
Trained on a bigger dataset. It is better at the "obscured camera" look but not as good at the changing motion. Added the trigger word CandidChange. Could still be good for stacking with other Loras for the "obscured camera" trick but I have not tested it much yet. For some things V1 may still be better but V2 should be better with front facing. I will try to make a merge of both if i can ever figure how to merge Hunyuan Loras in Comfy. Here is the random prompt I have tested with using RandomPrompts node:
CandidChange. A woman is {dressing|undressing|adjusting her clothes} in a {bathroom|locker room|bedroom}. She is {partially naked|wearing a bra and panties|wearing a casual outfit}. The camera view is {partially hidden by a stucture|filmed through a hole in the wall|taken from an open doorway|seen through a window|obscured by a foreground object}.V1 -------
Candid videos of dressing and undressing. There is no specific trigger, specific descriptions seem to work well. "A woman is getting dressed" or "A video of a woman changing her clothes" followed by descriptions of the video quality like "The camera is obscured" or "the video was taken from behind an obstacle."
Description
V2 -------
Trained on a bigger dataset. It is better at the "obscured camera" look but not as good at the changing motion. Added the trigger word CandidChange. Could still be good for stacking with other Loras for the "obscured camera" trick but I have not tested it much yet. Here is the random prompt I have tested with using RandomPrompts node:
CandidChange. A woman is {dressing|undressing|adjusting her clothes} in a {bathroom|locker room|bedroom}. She is {partially naked|wearing a bra and panties|wearing a casual outfit}. The camera view is {partially hidden by a stucture|filmed through a hole in the wall|taken from an open doorway|seen through a window|obscured by a foreground object}.FAQ
Comments (6)
It makes literally no sense to do this- and confounds the AI to a purposeless degree. Instead use 2 layers, and merge them in any of an endless number of editors, or do the same in a workflow.
Or, and hear me out, you could train an AI to do the same job with fewer layers and a simpler workflow. Sure, you could do it yourself through other means. Modern "AI" isn't exactly about doing it yourself. To me, it makes no sense to have ChatGPT compose an email to your mom when you could, you know, write an email to your mom. But people do it anyway. If the crux of your statement is that there is something lazy about training an AI to do something you could have done yourself, I think you might not be prepared for the robot apocalypse. (Apologies for the antagonistic tone, I was just a little floored by this hahaha).
@CZTehTron nah I get you, making this a LoRA is just way better / faster compared to post editing (better lighting, color accuracy and so on...)
Why make a PB&J when you can just eat the bread, spoon some PB in your mouth and then squirt jelly all over yourself?
@blobby99 Despite also agreeing with the PB&J comment... I think you aren't wrong. I wanted this LORA at first for a SFW generation of peaking in on an animal, so a peak/spy LORA and a changing LORA would be good, this is combined, also for good reasons probably. But now I want a peak/spy only.
Can you make a WAN 2.2 lora of this?