CherryBoyHunter_Waifu_OC_Style

Cherry Style V2 — Face and Eye ExperimentCherry Style V2 is an experimental refinement trained with greater emphasis on faces and eyes. Its goal was to improve facial definition while retaining the recognizable Cherry aesthetic.Because the V2 dataset was more selective, the LoRA has a stronger and less neutral influence. It may exaggerate glossy skin, highlights, blush, facial expressions, eye shapes, and recurring visual patterns from the training material.V2 should be treated as an alternative for face-focused images, not necessarily a direct replacement for V1.Recommended Settings<ul><li>LoRA strength: <code>0.35–0.65</code></li><li>Resolution: <code>832×1216</code>, <code>960×1440</code>, <code>1024×1024</code>, or <code>1024×1536</code></li><li>CFG Scale: <code>4.5–6.5</code></li><li>Sampler: <code>DPM++ 2M SDE</code></li><li>Scheduler: <code>Karras</code></li><li>Steps: <code>24–35</code></li><li>Batch size: <code>1</code> while testing</li></ul>Suggested UseV2 works best when you want:<ul><li>Portraits and upper-body compositions</li><li>Greater emphasis on eyes</li><li>Stronger facial expressions</li><li>A more concentrated Cherry Style appearance</li><li>Experimental close-up generations</li></ul>Start around <code>0.45–0.50</code> LoRA strength. Avoid stacking too many tags such as:<pre><code>glossy skin, shiny skin, wet skin, oily skin,
extreme blush, reflective skin, dramatic highlights</code></pre>These terms may amplify the glossy or “cooked” skin effect. For more natural results, try:<pre><code>natural skin texture, subtle pores, soft matte skin,
controlled highlights, diffused lighting</code></pre>If the face becomes distorted or overly stylized, reduce LoRA strength before changing the sampler or adding more negative tags.

Files with SHA256: e6e1b6c905c5cb0008f88dcba55629e00dfa1ecd7f6a82b7e0749f89b7434a38