MY pixiv link:<a target="_blank" rel="ugc" href="https://www.pixiv.net/users/76094070">https://www.pixiv.net/users/76094070</a>Changed to a new training parameter to train a LORA and simply generated some graphs. The results look good.The characters are more detailed.I hope you can download it to compare.<img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9996b691-1030-48a3-3014-9ccda3874b00/width=525/9996b691-1030-48a3-3014-9ccda3874b00" />This time I trained the model for clock departure, and here are some comparison graphs. I do not recommend using the EULER A sampler.<img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/4abcf709-a771-4bfc-960d-a6f5129f7500/width=525/4abcf709-a771-4bfc-960d-a6f5129f7500" /><img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0af3b50c-adb7-4596-e04e-f1f06af6d700/width=525/0af3b50c-adb7-4596-e04e-f1f06af6d700" /><img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/b7bb2579-bc24-46cc-7d87-5719313f8100/width=525/b7bb2579-bc24-46cc-7d87-5719313f8100" />

Files with SHA256: aab8a4203ae51984e04aa6f6036233750d8aba0dac307878e0d095fb21f9796b