Lora for drawing the folded pose.
Recommended weight 0.6-0.8
Trained on AOM2 (Also works fine with AOM3)
The result can be influenced by tags like spread legs, knees to chest, legs up, legs over head, leg hold, .
Important: often generates censored images, so use the censored tag in your negative prompt.
This time, I was much more careful about preparing the dataset, and now cropped images are rarely generated.
Description
learning
Tool: Kohya's GUI
73img * 100steps / batch size 2
FAQ
Comments (24)
Your trigger words contain <LORA:Folded:0.7> but the model download is called foldedpose_foldedposev1, this make it so the previously mentioned trigger word raises a 'lora not found' error when used. You might want to fix that
also, your preview images do not contain the trigger words
I see what you mean, seems like they might not be necessary?
Yeah, it just works, no trigger words necessary
Thanks, i fixed it
CivitAI may change the intended file name as part of their submission process. You can trigger any LoRa by using the exact file name.
@IShadowStar for me it turned out to be news, it turns out that I had previously indicated the wrong trigger word in all loras
and for some reason, using examples, it writes prompt and other information from img2img, but writes that it’s like txt2img, although this is not so
Hope for a example prompt for t2i,thanks
masterpiece,best quality,CG,wallpaper,HDR,high quality,high-definition,extremely detailed,1girl, nsfw <lora:foldedPose_v1:0.7>, folded
Negative prompt: (worst quality, low quality:1.4), censored
Steps: 75, Sampler: DPM++ SDE Karras, CFG scale: 6, Seed: 1968172149, Size: 480x768, Model hash: eb4099ba9c, Model: AOM3A3, Clip skip: 2
TXS!
hello, very good, your loRA wanted to know how you do it, do you have a link where you learned from, thank you very much
@Chik Thanks you are my idol haha
@Chik Hi, how you train specifically a position? I have doubts about the tags you are using when training
@VKTralala I just collected a dataset with images of this pose in each one indicated a folded token and tried to use any garbage tokens as little as possible
alas, I no longer have the datasat
It performs not good at aom3a1b, always take some mosaic at pussy. I love this lora but cant use it. Pity.
problem solved, put the trigger "folded" at first, and <lora: xxx> at last
try adding censored to the negative prompt
@Nyeilim It works! Thx.
@Chik i would love to use this but i have issues with it as well on ghostmix. can you please retrain it using NAI checkpoint as the base? that ensures that best compatibility with other loras and models
@icingqueen Sorry, but no, because the database has long since been deleted.
hello! may i know if there's a link to download the model from the cover/first picture? thank you!
Details
Files
Folded.safetensors
Mirrors
foldedPose_v1.safetensors
foldedPose_v1.safetensors
foldedPose_v1.safetensors
Folded.safetensors
foldedPose_v1.safetensors
foldedPose_foldedPoseV1.safetensors
foldedPose_v1.safetensors
Folded.safetensors
Folded.safetensors
foldedPose_v1.safetensors
foldedPose_foldedPoseV1.safetensors
foldedPose_v1.safetensors
foldedPose_v1.safetensors
foldedPose_v1.safetensors
foldedPose_v1.safetensors
foldedPose_v1.safetensors
foldedPose_v1.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.