So V1 has alot of bdsm and a big data set 950 pics , v2 is morer trimmed down , no bdsm version 2.1 is a tweak of v2 fixing some face distorts on some models ..*******Important info*****If you are having issues with the keyword triggers, try using just the filename of the Lora ex SybianV21 if your using 2.1

Sybian LoRa

Getting some feed back about the face rendering, this was 8500 steps on about 100 Hd photos, for training it was 512x512 not sure how this will do on larger photos, should be fine . i took out all the moaning, and used this mini 2.1 version, to focus on fixing gthe face issue in 2.0 , and also getting good clean view from behind, with a data set that tried to fix the isues. I had to scrool through quiet a few videos and pause and sceen shot certain times. Keep the feedback coming in comments so on teh bigger versions, I can dial in good data that works. The main trigger is Sybian, but if using WEB UI (automatic111) or you are having issues try using SybianV21 instead ...

Files with SHA256: f52f8d2784cd0601cb1efaa707c7159dfeb6c918d3b214b6e253b87be5b3bfdf