💡 提示 / Note

建议使用 int8 RDBT 基础模型进行加速。也可以配合 RDBT LoRA 模型进一步提升速度，但RDBT LoRA可能会使原人物的画风产生偏差（效果因人而异）。此外，使用美学版模型时无需添加评分标签（如 score_7 等）。

It is recommended to use the int8 RDBT base model for acceleration. You can also use the RDBT LoRA model for speedup; however, RDBT LoRA  may cause slight style deviations from the original character design (results may vary by personal preference). Additionally, quality tags (e.g., score_7) are not required when using the Aesthetic version model.

Anima Version

🔧 Recommended Settings

• Weight: 0.8 - 1

• Base Model: Anima int8

• Lora Model: RDBT | Anima rdbt_v1.0_anima_b1_16-step.safetensors

• Steps: 16

• CFG Scale: 1.0

• Sampler: Er_sde

• Scheduler: Simple

🔍 Hires Fix (Upscale)

• Denoising Strength: ≥ 0.4

🌟 Quality Prompt

masterpiece, best quality, anime coloring, anime screencap

🧩 Basic Prompt

• Prompt :

Jae-Eun,brown hair,long hair,brown eyes

🚫 Negative Prompt

worst quality, low quality, artist name, jpeg artifacts

Illustrious Version

🔧 Recommended Settings

• Weight: 0.7 - 0.9

• Model: WAI-NSFW-illustrious-SDXL

• Steps: 25 - 30

• CFG Scale: 7.0

• Sampler: Euler A

• Scheduler: Normal

🔍 Hires Fix (Upscale)

• Denoising Strength: ≥ 0.4

🌟 Quality Prompt

masterpiece,best quality,very aesthetic,absurdres,4k,8k,intricate detail,amazing quality,anime coloring, anime screencap

🧩 Basic Prompt

• Prompt :

Jae-Eun,brown hair,long hair,brown eyes

🚫 Negative Prompt

bad quality, worst quality, lowres, bad anatomy, jpeg artifacts, ugly, poorly drawn, signature, watermark, username, blurry, 4-koma, sketch, bad, error, fewer, extra, missing, bad quality, watermark, unfinished, displeasing, chromatic aberration, signature, extra digits, artistic error, username, scan, abstract, blurry, lowres, low quality,big head, bad anatomy, artist name, watermark, weibo username, derivative work, one-hour drawing challenge, jaggy lines, female pubichair, address, patreon logo, twitter username,sound effects, futanari, furry