Well. I ported another existing LORA that I made, this time a Japanese pink-haired VOCALOID virtual singer in chinese dress outfit. Same existing datasets, nothing unchanged, just I add some captions that is not AI written and simplified so you can just remember a couple of phrases and no trigger words are needed, you have to make sure that the LORA is loaded so it can parse some of the keyword phrases to create its likeness from the LORA. Works well with most of ZIT-based models.Second version uses completely different datasets but still carries same exact theme. Includes trigger words to get image generation accurately. <h3 id="prompt-examples:">Prompt examples:</h3><pre><code> a woman in red chinese dress and a very long pink hair standing with alluring pose. She wears fishnet legwear and a red ribbon on her hair.</code></pre>Other prompt examples can be checked on its sample images here. Edit out when necessary. <h3 id="generation-settings:">Generation settings:</h3><ul><li>Sampler: Euler or UniPC er_sde or any equivalent that is working</li><li>Steps: 8-10</li><li>CFG: 1.0</li><li>ComfyUI is recommended</li></ul>

Chinese Luka ZIT LORA

Second version. Uses completely different datasets but theme remain the exactly same.

Files with SHA256: 620cc3d1b851dd782f17ffef9e373bb2a1045d105b3b59e5a515fbc598519c19