This is a lora trained on a Sexy dance and twerking, part of dataset have the same music so it may have a loud impact on music in the backgound, you can prompt to manage that with non_diegetic_music and overall_soundscape, It's a T2va lora but it will probably work on i2va too even in ref2video in some of my tries ( using hybrid model)
If you need more nude anatomy you may need a lora for that as it was trained only for motions with 5 sec trimmed video , it's very expensive in compute with this duration but captionning dance under was a bit tricky..
Please post your video , love to see other people works :)
All video posted are exclusivly T2v
Base model is already capable of really sexy dance , this lora may boost the "sexy" thing ;)
exemple of Prompt in my dataset for inspiration: (edit post all my captions :) )
integrated_multimodal_description: [Shot 1] Live-action, close-up framing. A young woman with long blonde hair (S1), wearing a white polka-dot crop top and light blue pajama pants with pink donut prints, stands in a bedroom. She has her hands in her pants pockets, then pulls them out while tilting her head back and forth. Behind her, gray curtains, a bed with white sheets, and a green-and-yellow cat tree are visible. At 00:00.000, she begins moving her hips rhythmically to electronic music; at 00:01.500, she adjusts her pants waistband downward to reveal her midriff. She continues dancing with sharp hip movements and hand gestures, then points toward the camera at 00:03.500.
overall_soundscape: The room is quiet except for the rhythmic thud of feet shifting on carpet, fabric rustling as the subject adjusts her pants, and soft breathy exhalations during dance moves. No wind, traffic, or environmental noise is present.
non_diegetic_music: Upbeat electronic dance track with a driving four-on-the-floor beat, synthesized bassline, and repetitive vocal sample loop ("now now now now") at fast tempo; volume remains consistent with no dynamic shifts.
If you interested in some prompts for video posted here just ask ;)
Description
FAQ
Comments (18)
when i see a sexy dance lora for a video model: instant download!
Hope you will have fun with this one, i love it, not just because i trained it :-p
@inuptia758 I love it so far! i really would like some text prompts for some of the example videos though since my prompting for minimax h3 is not the best lol!
@TypicalLoraLover Just posted all of them, exept those i created by myself in optional files ;)
@inuptia758 ahh thanks! I've been having so much fun so far with this!
Pretty awesome. Would actually love to know what your prompts were..
i let my captions on optional files , it may help you , i used to load all my caption to a llm in order to craft some new one
@inuptia758 actually found the metadata thanks :) by the way, is it only T2V ?
@hboxgames132 did not test
Fantastic work!!
a man of culture hats off
might actually be my favourite H3 lora. unlimited fun with this one, great job
Glad you like it :)
Works great even with i2v. It's heavy on the music, so perhaps in a retraining you could strip the audio out of the files and training without?
yeah, like i said on the description, i think i put too much video with the same music maybe 1/3 of the dataset, i think the model just understand it have to learn that by redundancy
The dataset caption is gold, thank you 🐸
yeah i think it's better than giving some trigger :)
i know i already commented on this lora, but I've been having SO MUCH FUN with this lora for real! my favorite lora for minimax h3 right now! I can't wait for a v2 if its ever planned! thanks for sharing this lora! I really love it!