This is for all the freaks out there. This lora can work as a detailer lora (if your detailer can detect tapering penises) or a main image lora. As a main image lora, I recommend lowering the strength to about 0.7 to prevent style bleed. When used in the detailer you can bump that weight up to 1 or higher. The trigger words are listed on the right, be sure to mix and match and add parenthesis if needed.
The dataset for this one is a bit all over the place tbh and contains quite a few images that aren't technically tapered but close enough imo. I tagged all the traditional tapered penises as "thin penis tip" so feel free to use that tag if that's what you want. I also struggled a lot with fdetailer on this one as it can't seem to recognize tapering penises as penises. If anyone has any suggestions on a good detailer for this feel free to comment.
There are 2 ways this lora can be used and both methods were used to generate the sample pics:
Use as a normal lora. Works but can change the style and composition of your image. Use a weight of 0.7 or lower to reduce that. You can always add a higher weight to the lora in the detailer later to bump it up in the details.
Use only as a detailer. Great way to boost the peens visuals with 0 affect on overall image style. I personally recommend this approach but I know its not available to everyone.
For those curious on how to do method 2, here's a quick guide.
Generate your base image the same way you usually would up until applying detailers. Leave out the lora or include it with a low weight, like 0.7.
Add trigger tags like "m0nster t4pering pp, genital slit, nubbed penis, spiked penis,, ribbed penis" if you want a specific look.
Plug this lora into either your adetailer or fdetailer.
Adjust the weight of the lora and the denoise of the detailer as you want. In my testing, a weight of 1 on the lora and a denoise of 0.40 is the sweet spot. The higher the denoise the higher the risk/reward.
If you're struggling, check the quick example workflows I have attached to the sample pics.
Description
Trained on the below params.
172 images:
{
"engine": "kohya",
"unetLR": 0.0005,
"clipSkip": 2,
"loraType": "lora",
"keepTokens": 1,
"networkDim": 16,
"numRepeats": 5,
"resolution": 1024,
"lrScheduler": "cosine",
"minSnrGamma": 5,
"noiseOffset": 0.1,
"targetSteps": 4300,
"enableBucket": true,
"networkAlpha": 8,
"optimizerType": "Prodigy",
"textEncoderLR": 0.00005,
"maxTrainEpochs": 20,
"shuffleCaption": true,
"trainBatchSize": 4,
"flipAugmentation": true,
"lrSchedulerNumCycles": 3
}
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.