Trigger words:
Spank (or any of its variations like Spanks, Spanking, Spanked, etc)Example prompt:
You can download any of the showcase videos and drag them into ComfyUI. They all have the metadata attached.
A gorgeous 25-year-old redhead woman with her hair in a ponytail and with her round butt towards the camera. She is wearing a a red bra and red lace thong. She is bent over a bed. She is being spanked. Her butt jiggles with each spank.
Realistic. Rear-view shot.TIPS:
This is being re-trained. it has only been trained for rear-view shots as of right now. It is still being trained, but is already more consistent than the previous version.
I will introduce other angles (Side-view, Front-view) in the future as training progresses. I'm hoping breaking it into separate training sessions helps.
Description
Alpha preview of the newer training methods.
Current training: 2000 steps at 1e-4, then 2000 steps at 2e-5
Currently only targeting Rear-View spanking only. Other angles will be added in the future.
Any variation of the word "Spank" should work.
FAQ
Comments (27)
Nice! let's try this new one. Humble request: "Own spank" ;)
I may need to dig for a dataset of that, but I can add it to the list. Finishing the current training, then extending the current version to include side-view is on the list.
@ComfyTinker Sounds great! I can help with the dataset... but I would need some guidance 🙃
@Xavis If you have a compiled list of videos with good clear self-spanking, you can DM me and I can provide some guidance.
@ComfyTinkerI mean I don't have them yet, but I can work on that during my daily calls lol. What length would work best? I'll DM you.
@Xavis Anywhere between 1-5 seconds. If you have longer, that's fine. I can trim, crop, and downscale where need.
@ComfyTinker Hey there! Question?¿ preferably with visible faces right?
@Xavis Faces don't need to be visible. It's actually easier to train if they're cropped out.
@ComfyTinker Oh! gotcha, yeah make sense, I'll send you some by EOD
haha you got the hotdog working :)
Works awesome, thank you
so you can add clips in the training data while resuming a training and it will pick up and add them automatically ?
I can start training from a pre-existing LoRA, yes. I can train all of these rear-view videos, then train a side-view later.
why people stop putting prompts on almost all hunyuan loras ?
Save video locally, import into comfy, prompt is embedded into workflow.
Because all of my videos have embedded metadata, so you can replicate each of the videos exactly.
@ComfyTinker Not everyone uses Comfy friend. A prompt is still helpful.
Do you want more spanking clips for your dataset? Bigger butts and more jiggles.
Can you do one for WAN 2.1 please?
I would also like to make a request for a wan 2.1 version also please
The action is smooth and realistic. Excellent. I'm learning about all of this and would this work in Krita with stable diffusion? would it take a still AI image and create motion? Can this work spanking a man?
You might want to take a look at the WAN spanking LoRA. It has a better dataset than this one and is better trained. You might be able to get Krita to do this, but I'm not familiar.
@ComfyTinker Thanks so much
This works in framepack-studio? I don't understand why mine doesn't work. I'm new at this and tried to make a "concept" lora for spanking using video clips. It was my first try and it doesn't work in FP-studio. It says Hunyuan 720 [fp8] .
I tried to use it and got an error the error below. Did I do something wrong?
Model doesn't have any LoRA adapters or peft_config.
[After unloading LoRAs] Transformer has no peft_config attribute
[After unloading LoRAs] No LoRA components found in transformer
It's possible I skipped over something crucial when I created it though. All I did was put in captions and blast through the rest. IT was very frustrating because when I uploaded my videos there were a bunch of dupes for no reason and when I tried to delete the dupes it would delete other videos and not the one I meant to delete. It seems very buggy.
1. Update FramePack, this sounds like an old bug
2. If don't have any other LoRAs then make a copy of this one in the same directory so it detects 2, I had the same issue before where it would only load LoRAs if it could find more than one.
Lol