Gaping ( )( )

IMPORTANT: This is NOT a SLIDER LoRA like all my other LoRAs. This is a normal LoRA trained on a dataset.

There seem to be a trend lately that creators try to make the larges possible LoRA with over Rank200 and over 1GB large LoRAs, but i dont think that is nessesary, and i wanted to test if that is really needed.
So i trained this 7mb Rank1 LoRA on a 25-image dataset. And i want to share the setting i used if anyone else want to make fun small LoRAs like this.

LR: 0.00002
Steps: 10000 (batch 1), but half or less would probably do as well.
Timestep Type: Linear
Timestep Bias: Low Noise
Loss Type: Wavelet
Dataset: 25 images with captions. No faces.
Res: 768x768 or similar
Wavelet has given me some problems on other LoRAs i have made when the dataset had one bad image out of a dataset of 30 images. So be aware that wavelet is probably not the most forgiving and best setting to use.
This was only intended as a testrun, since i dont train normal LoRAs at all. So the dataset could probably have a bit more variation on a V2, if i decide to make a new version. But i think it turned out pretty good for what it is.

Only females in the dataset.
But it should be possible to use with male/trans at lower str.

I used a trigger word, g4pe, as well as natural language captions. But i dont think the triggerword does anything significant. You are better off just including gaping with natural language in the prompt.

Recommended range: 1.0 to 1.8

(but it depends if used alone, with Bypass, or other LoRAs etc)

Can be combined with all other LoRAs without any problem. And even though it can go way past 1.0, i dont recommend pushing past 1.5 on closeup, and not past 2.0 on wide shots.