This is my take on Serial Experiments Lain LoRA. It is trained with alternating checkpoints including:<a target="_blank" rel="ugc" href="https://civarchive.com/models/129681/sdxl-faetastic">https://civarchive.com/models/129681/sdxl-faetastic</a> (2.4)<a target="_blank" rel="ugc" href="https://civarchive.com/models/84040?modelVersionId=309729">https://civarchive.com/models/84040?modelVersionId=309729</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/127923/sdxl-yamers-realistic-5">https://civarchive.com/models/127923/sdxl-yamers-realistic-5</a> (V5 + RunDiffusion)<a target="_blank" rel="ugc" href="https://civarchive.com/models/118756?modelVersionId=339195">https://civarchive.com/models/118756?modelVersionId=339195</a> (v33515)<a target="_blank" rel="ugc" href="https://civarchive.com/models/267728?modelVersionId=345685">https://civarchive.com/models/267728?modelVersionId=345685</a> (FusionOG)and some others. Thus, this LoRA may work with different checkpoints, although anime-friendly checkpoints can give better results because of the nature of images in the training set.Use "Iwakura Lain" for activation.The braid can be generated at wrong side, but flipping the image horizontally helps

Files with SHA256: aa4f854f141bcac1b398da5755a67a36082f4cf1aa7ef1c1f825e5a362b4fb0e