Motion of Waist Dance.
Can use with characters LORAs
Example of this LORA in Video Mix
One Click Generation: Wan2_2_WaistDance_T2V_14B
Remember to change the character LORA and prompts.
Experience link: https://www.runninghub.ai/ai-detail/1955436186233966593/?inviteCode=rh-v1005
Prompts Example:
[Shot 1: Medium shot, full-body shot, eye-level perspective, flat focus. The subject is a young woman dancing in a dimly lit dance hall. She wears a short, red, midriff-baring skirt. Her waist sways dramatically to the rhythm, her hips swaying in time. Her hands are naturally raised, fingertips lightly tapping her shoulders, her long hair flowing with her movements. The background is a backdrop of flashing neon lights and blurred crowd silhouettes. Her expression is charming and slightly lazy, her eyes wandering, the overall atmosphere warm and sexy. Warm-toned spotlights illuminate her dancing figure, and the hem of her skirt is shimmering.], Dissolve, [Shot 2: Close-up shot, from the waist up, over-the-shoulder perspective, telephoto lens. The subject continues her dance, revealing her slender waistline as she twists her waist. Her abdominal muscles tighten slightly with the movement, her neck tilted back slightly, her hair falling across her chest, a faint smile playing on her lips. The background is blurred with the lights of the dance floor, creating a seductive atmosphere. The sidelight casts soft shadows on her cheeks and collarbone, highlighting her curves.], Waist Dance
Guys, I've discovered a treasure product for AI images and videos called RunningHub! Global ComfyUI developers release hundreds of extremely interesting and useful AI applications here every day, covering both entertainment and work aspects.
Register link: https://www.runninghub.ai/?inviteCode=rh-v1005.
Register and get 1000 RH coins, and you can generate a large number of images and videos for free!
Description
High noise
FAQ
Comments (16)
i2v can use it?
No, I think. no debug will show, but also no use at all.
Can you make a version without the zoom?
What about the low noise model? No necessary or do we need to plug the same HN model into the LN flow?
Also want to know this :)
Haven't tried this lora yet, but from how the 2.2 high/low models work in general, the high does movement/composition, and the low models refine details. In this case, the high model should already have locked in the movement (the dance), a low model is not needed as there's no details to paint in.
@digos132 right answer, digos132 is right
you can use any low noise model, my suggestion is some quality enhence ones are fine
If it helps, at the end, I used this lora as "low" model too and along the Lightning lora, I had pretty mixed results. The "slow mo" issue of the Lightx2v made me to investigate a lil bit more and then find the idea of the "3x Sampler". So the 4th video of my last post here shows it in action. And BTW, used this lora in all the subflows (with different weights).
it depends on the model, but generally speaking you don't need the low noise model, but if you do have both, the low noise helps refine the movement details during the render pass... but with that said, a/b comparisons often show little improvement. I rent gpus, so I spend a few hours at a time troubleshooting/testing shit like this. And all this is to say that if you DO have a low noise, use it, but its not required in some cases, AND, I've gotten more than a few 2.1 loras to work on the high noise 2.2 with lightning loras. And yes, it seems the lightning 2.2 loras work best when both are used.
@Karlmeister_AR where is the workflow?
@panjool Embedded in the videos
More dance Loras would be great. Thank you for this one. May all your AI-generated works be fast and of good quality. Thank you.
can you make a 'magic bomb' dance lora?
if you dont know what it is search google:
magic bomb dance challenge