This lora adds a man's hand in POV (point of view scenario) to an existing image in editing mode.
it also works in text2image mode, use same prompts.
I was was trained on 30 pairs for 8000 steps in oneTrainer on base checkpoint. Use with distilled checkpoint with 4 steps.
prompt template: POV hand, zoom in, man's left hand is grabbing her boobs.
hand can: touch, grab, pull.
There were some nsfw pics in a dataset, but you'd better add some other nsfw lora for that purpose.
Description
FAQ
Comments (7)
f2k_hand_pov_8000s.safetensors' contains unexpected module keys, it is likely that this Lora is for a different model : '['double_stream_modulation_img.linear.lora_down.weight', 'double_stream_modulation_img.linear.lora_up.weight', 'double_stream_modulation_txt.linear.lora_down.weight', 'double_stream_modulation_txt.lin...'"
i think onetrainer adds some additional keys. Lora works ok in comfy.
@tensorbanana Ah fair enough. tried it on LTX2 in Wan2gp ui
Yeah same problem with wan2gp with few more loras. I dont know how to fix it
Amazing model, keep it up, would love to see more updates on this
In future versions, I would love to be able to do this on a particular person on an image with 2 or more people in it (like do this on person on left / right / this color cloth / that jean pant person). Right now it gives deformed results on 2 more more people.
Sick lora. Who knew a hand would be so fun
Details
Files
f2k_hand_pov_8000s.safetensors
Mirrors
f2k_hand_pov_8000s.safetensors
f2k_hand_pov_8000s.safetensors
f2k_hand_pov_8000s.safetensors
KLEIN_HandPOV.safetensors
f2k_hand_pov_8000s.safetensors
f2k_hand_pov_8000s.safetensors
f2k_hand_pov_8000s.safetensors
Flux2Klein9b_POV_Hands.safetensors
f2k_hand_pov_8000s.1.safetensors
f2k_hand_pov_8000s.safetensors
f2k_hand_pov_8000s-mid_2347738-vid_2640656.safetensors
f2k_hand_pov_8000s.1.safetensors
f2k_hand_pov_8000s.safetensors