signalviolet_style — Anima LoRA (Style)
signalviolet_style is a style LoRA fine-tuned from the artwork of the illustrator signalviolet, capturing their characteristic soft, delicate anime aesthetic — clean cel shading, gentle blush, and cute, expressive heroines with a slightly glossy finish.
Training details
Base model: Anima Base V1.0 (circlestone-labs)
Dataset: 327 curated illustrations (danbooru/Gelbooru style tags)
Network: LoRA, rank 32 / alpha 16
Epochs: 10 (~820 steps), AdamW 8-bit
Architecture: anima-preview/lora (SDXL-based, epsilon prediction)
Trigger word
Plain Text
signalvioletPlace it at the beginning of your prompt (kept token).
Recommended settings
ParameterValueBase modelAnima Base V1.0Text encoderqwen_3_06b_baseVAEqwen_image_vaeLoRA weight0.6 – 0.8 (start at 0.7; lower for complex compositions)Sampler / Schedulerer_sde / simple (or euler_a, dpmpp_2m_sde)Steps30 – 50CFG4 – 5Resolution512² – 1536² px (e.g. 832×1216, 1024×1024)Recommended positivemasterpiece, best quality, score_7, safe, signalviolet, ...Recommended negativeworst quality, low quality, score_1, score_2, score_3, artist name, blurry, jpeg artifacts, chromatic aberration
Description
Initial release. Trained on 327 images with Anima Base V1.0.