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, 13 epochs of training

Files with SHA256: 336db673b4b98dc066520ddde8eb0badfe3daf4166dfc9e94be7d7e5f5fb97f4