<h2 id="new-version-3.0.-k5o6hs2tu">New Version 3.0.</h2>I'll give a brief summary of the differences between the various versions.<ul><li>Version 1 was created with a larger image dataset (about 40-50 images), but without cleaning or resizing the images. It creates more "fantasy-like" environments, but it's also easy to encounter horrors and quite disappointing results...</li><li>In version 2, I reduced the dataset to a few images (about ten), but all were cleaned. The results are more limited, but clearer.</li><li>With version 3, I slightly expanded the dataset (about 20 images), but each image was generated using img2img combined with LORA, similar to this, to train the LORA with more defined images.</li></ul>I'm not sure if this can be done for previous versions of this LORA, but to drastically improve the results, I generated the images posted here using img2img by uploading monochromatic images (white, pink, green, blue) depending on the background + Denoising strength set to 0.75. For example, with a pink background, a tentacle pit will automatically be generated!!

Files with SHA256: 881704f176a2da70c7eea3c15432b53d2da6cf6430d303e49fd376deb425a87c