Rather than using HameF, which is compatible with Anima 2.9B… I’ve retrained the model from scratch.

The MOSS-image version is still in beta. To be honest, MOSS-image’s LoRA capacity is so high that I can’t quite keep it under control. I’d like to prepare a wider variety of images, but I’m hesitating to boot up SD 1.5 at this stage.

positive prompt: masterpiece, best quality, amazing quality, newest, cute, digital art, (One point eyes highlight:1.2),

negative prompt: worst quality, low quality, normal quality, signature, username, colored pupils, gradient eyes, glowing eyes, two-tone eyes, blurry eyes, bad iris, logo, bad anatomy,

Anima2.9Bに対応したHameF・・・ではなく、新規に学習し直しました。

MOSS-image版はまだベータです。正直、MOSS-imageのLoRA受付力が高すぎて制御しきれません。もう少し様々な画像を用意したいのですが、いまさらSD1.5を起動するのもなあと二の足を踏んでいます。