<h1 id="v4.3-final">v4.3 [final]</h1>I recommend to choose this one or v2. If you don't care about speed and want great anime results use this model + LCM sampler + 20 steps. It's simply better than any model I know in this aspect.<h3 id="recommended-settings:">Recommended settings:</h3>6+ steps (8 + 4 steps of upscaler is optimal for my use cases, feel free to experiment).A few options of samplers:<ul><li>CFG 1.8 | Euler A</li><li>CFG 2.0 | LCM - better for anime and other abstract things.</li><li>CFG 1.8 | DPM++ 2M SDE with BETA scheduler (normal doesn't work)</li></ul>Use style tags such as "anime style, flat coloring" or artist/title name. It will help model to know what you want from it.The fastest generation: use CFG 1.0 and set lora CFG_slider on negative -0.4 value. Lora listed in suggested resources. (CFG 1.0 ignores negative prompt. Don't use it if you need negative)The best coherency: use CFG 2.8 and CFG_slider lora on +0.75 weight (for Euler A).<h1 id="v3.2b">v3.2B</h1>Recommended settings:<ul><li>CFG 1.8</li><li>LCM / [kl optimal / sgm uniform]</li><li>5-10 steps</li></ul><h1 id="v3.2a">v3.2A</h1>Recommended settings:<ul><li>CFG 1.8 (works not bad on cfg 1 as well)</li><li>Euler A / normal</li><li>5-10 steps</li></ul><h1 id="v1.99-(v-prediction)">v1.99 (v-prediction)</h1>Use title or artist name to get a specific style.Eauler A - Beta scheduler (or normal)8-16 stepsCFG: 1.0 recommended for 2x speed. Use more if you want to use negative prompt.<h1 id="v2">v2</h1>Sampler: Euler A / Euler A cfg++CFG: (depends on sampler)Alternative sampler: LCM

ALF mix (IL, 8 steps generation)

For this model were chosen the best blocks between v2 and v3. It can operate with dark colors, have ~medium randomness of content (I can't do it bigger if I want to save good coherency). The model doesn't contain extra loras for quality. So it will not conflict with your lora improvements.

Files with SHA256: 3f7f84bfe91cf9be4b508200b755dcc4805e616044202b525fe8d7ddd81e301c