<h1 id="heading-2442">Kohaku V5</h1><h2 id="heading-2443">Details</h2>I trained a style LyCORIS model with Kohaku V2.1 as base model. And then merge it into kohaku V3 rev2 and Kohaku V4 rev1 as Kohaku V5 rc2/3. Finally I found 2 better merge ratio between rc1 and 2. Which becomes rev1 and rev2.<ul><li>Kohaku V5 rc1<ul><li>Base Model: Kohaku V3 rev1</li><li>LyCORIS weight: 0.4</li></ul></li><li>Kohaku V5 rc2<ul><li>Base Model: Kohaku V3 rev2</li><li>LyCORIS weight: 0.4</li></ul></li><li>Kohaku V5 rc3<ul><li>Base Model: Kohaku V4 rev1</li><li>LyCORIS weight: 0.2</li></ul></li><li>Kohaku V5 rev1:<ul><li>rc2 + rc3</li><li>Weighted Merge Ratio: 0.5</li></ul></li><li>Kohaku V5 rev2:<ul><li>rc2 + rc3</li><li>Weighted Merge Ratio: 0.65</li></ul></li></ul><h2 id="heading-2444">Usage</h2>Just use it with danbooru tags, use EasyNegative or EasyNegativeV2 as Neg emb. No need to use lot of neg prompt or quality tags.

Files with SHA256: b841998f4bd69bea4520201b7b3b046b458633629b979ebc30076d5affe334bf