<h3 id="usage-tips">Usage Tips</h3>This model was trained on Anima base v1.0.V2 is a refreshed version of style LoRAs I made when I was still a beginner, rebuilt with more stable training settings.Trigger word: Use the version file name as the trigger word. Examples: <code>@qr4k</code>, <code>@fr8t</code>, <code>@fr9t</code>This is a collection trained primarily on styles frequently used in NAI.Versions with different names were trained with different datasets and settings.Please refer to the sample prompts as a basic guide. Depending on the image, using a different prompt may produce better results.<h3 id="test-settings">Test Settings</h3>These settings are not required. They are the settings used for the sample images.CFG: 4–5 Sampler: er_sde Steps: 30–40 Highres: The sample images were generated at 1024×1536 resolution with 40 steps, then upscaled with 1.25× Highres at half the main steps. Highres is recommended to preserve the model’s distinctive style and achieve sufficient detail.

Anima korean NAI artists style

anima_qr4k_v2.0-epoch24.safetensors

qr4k_v2.0

Files with SHA256: 15d64b2e4bba037c0e87ca31d1aee04f6d9a3d9c84d02caa4c5faf44ed74f7d9