<h1 id="heading-2758">Cheesesteak Functions</h1>Use cheesesteak to activate. You can use prompts like bokeh, from above, and food photography for better angles. It also likes the words gloss and smoke if you wanna make it look extra delicious. Clip skip 1 yields better cheesesteaks, clip skip 2 yields what I'll call "out of town cheesesteaks". Best used with (and trained on) the Neverending Dream checkpoint. Trained with a 25 image dataset. As usual, I use Comfy. You can get a visibly okay cheesesteak without upscaling, but I suggest at the very least upscaling. Hires fix works great too. There is human data that can be summoned as well. Summon humans at your own peril, they're highly untested. Captioning is done with WD14, mostly because I'm not good at more advanced captioning yet and can be more consistent this way.<h1 id="heading-2759">A Word On Cheesesteaks</h1>Despite my better judgment, I allowed cheese wiz data to exist in the training prep. This is because I know that there are sickos out there who actually like eating cheesesteaks with cheese wiz. While I deeply loathe the depravity of such a cheesesteak, I am an accepting person, so you will occasionally spawn cheese wiz. Maybe in the future I'll make a new version where you can prompt it. The lora is not called "Philly cheesesteak" because only out of towners call it that.

Files with SHA256: 9d76236b0d32f1d76b3838a727c12824dcea645357d879b529aa634985e50898