First try at a Medieval Drolleries LoRA: SDXL, adafactor, trained as a styleFor the moment the only version is practically unusable, probably for insufficient cooking time: it's so weak it has barely any effect, and even without any filter the output isn't even near the originals

Medieval drolleries XL

I fear this is sadly undercooked. I underestimated the necessary epochs, probably because I didn't consider the relevant differences in style from the base models

Medieval_drolleries.safetensors

Second try at the medieval drolleries LoRA.This time the 300 images are processed for IL with Prodigy and a lot of epochs. Theoretically as a concept, but I still do not understand what differences there are between that and styles...It is possibile the tags 'parody' and 'fine art parody' will get you some result, those were chosen by the automathic tagger.For the rest, you have to softly alter the style. Do not blindly use embeddings, some tag could ruin the soup. These work quite well: <span style="background-color:rgb(37, 38, 43);color:rgb(140, 143, 163);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:14px">fine art parody, faded background, basic coloring, colored ink \(medium\), branches, inkblot, leaves, text, soft gradients, detailed, stylized anatomy, side view, antiqued, rough paper texture \(traditional media\)For some negative, reduce 3d and complexity: <span style="background-color:rgb(37, 38, 43);color:rgb(140, 143, 163);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:14px">volumetric shadowing, 3d, render, complex background, photorealistic

Medieval drolleries

First try at a Medieval Drolleries LoRA: SDXL, adafactor, trained as a styleFor the moment this is practically unusable, probably for insufficient cooking time: it's so weak it has barely any effect, and even without any filter the output isn't even near the originalsReally, try something else

Files with SHA256: bfb611c88a58ff31c7d8fd77276a6a195123ea0bdcbeee89afb190edd6fce687