Disclaimer: V1 / V2 are recommended for offline generation due to instability.
This LoRA works for both SFW and NSFW outputs. Please see individual version details for triggers (they are different per version) and specific notes.
Thanks to all who paid for access to V2, I am able to release V3 free for all. Tips are however appreciated as they pay for further training attempts to create new versions of this LoRA and others. Currently, this is the final revision planned unless I get an influx of buzz to support training; however, V3 is already great quality based on testing so long as you are using MiaoMiaoHarem Anima 1.6. I'm sure other models work, but this has been my go to for a while.
Description
Significant improvement across the board. Better framing for voyeuristic shots and there is better balance between the concepts. I've addressed a lot of the instability of V2 and V1, but it is not completely gone.
the POV triggers are intended to be combined with voyeurview. they all work well, but there is some instability on pov window. be mindful of image framing and you should not have a problem. if you are using full body on a voyeurview concept it could cause character limbs to stick through walls depending on the rest of your prompt. the two are not incompatible, but it does make it more likely the output diverges from expectations. another issue with the pov window is occasionally it will put the viewer reference inside and the subjects outside. it can make for interesting outputs, but they usually look wrong and will require cleanup.