Theobrobine is trained with 80+ theobrobine's videos, with its own time cropping, basically between 5s-10s, nearly 50,000 steps, with a loss of about 0.058, and is now using natural language tagging, and manually proofreading the accuracy of the tagging, which is a bit better than the 1.0 version.
Keywords: theobrobine
Recommended weight: 0.8
Cue word example:
This is a digital video taken from a close-up camera angle of a young woman with long flowing silver hair and stunning red eyes. Her skin looks extraordinarily fair in the soft, warm light, and she gazes confidently at the viewer with a slightly shy, serene expression. Her slender, curvy body was accentuated by her ample breasts. The plump breasts were being breast-fucked by a huge cock squeezed between her hands. Long hair cascaded over her shoulders with subtle green and white highlights. A vibrant gradient background of oranges, yellows and purples adds movement and intimacy to the scene, making it an exquisitely detailed scene.
theobrobineDescription
FAQ
Comments (4)
can you make one for WAN aswell
wan2.1 t2v doesn't seem to be able to generate NSFW content, I've tried training it and it works very badly, but training SFW content with the same parameters is no problem, so I don't have any good ideas at the moment
@Jarcis thanks for letting me know :)
@dex345 It may be my problem, as there is indeed a NSFW Lora model as well, will try again at a later stage





