Different merges seem to not have enough weights for the concept of this specific rendition, so here's a LoRA.Tested with the weight value of 0.8. For better results try adding<pre><code>cryptiacurves \(modeler\), da.nilkaz \(modeler\)</code></pre>to the positive prompt.Wasn't sure which epoch looked better so there are two versions available:<ul><li>Epoch 10 - may deviate from the original rendition</li><li>Epoch 13 - a tad closer to the original rendition, lighting seems to be better, but could be overtrained</li></ul>Might do other depictions of the character in the future if the stars align.

SCP-1471-A - MalO (cryptiacurves) - Illustrious

Initial dataset, 10 epochs of training

Files with SHA256: 7a47dc85c35573c6fe544d7a2e154ef5252ef4fa3d27ab0c3b2b15b3336b16d2