MiaoMiao HRPF 是一款经过精细训练的高质量风格 LoRA 模型，旨在为生成画面带来更加稳定、自然且具有细节表现力的视觉效果。
本模型在人物塑造、画面氛围、色彩层次以及细节控制方面具有良好的表现，能够有效提升图像的整体完成度与风格统一性。它适用于角色立绘、半身像、插画封面以及多种具有氛围感的艺术创作场景，在保持主体清晰度与构图稳定性的同时，为画面注入更精致的风格表现。
由于该模型的风格影响相对集中，建议在较低权重范围内使用，以获得更自然、平衡且稳定的生成效果。
MiaoMiao HRPF is a finely trained high-quality style LoRA designed to deliver more stable, natural, and detail-rich image generation results.
It performs well in character rendering, atmosphere building, color layering, and detail control, helping improve overall image polish and stylistic consistency. It is suitable for character portraits, half-body illustrations, cover artwork, and various atmospheric creative scenes, enhancing the image with a refined stylistic finish while preserving subject clarity and compositional stability.
Because the stylistic influence of this model is relatively concentrated, it is recommended to use it within a lower weight range for more natural, balanced, and stable outputs.
⚙️ Recommended Parameters / 推荐参数设置
| Parameter (参数) | Value / 设置值 |
|---|---|
| LoRA Weight <br> (LoRA 权重) | 0.6 <br> (不建议超过 0.65 / Do not exceed 0.65) |
| Sampler <br> (采样器) | Euler a / DPM++ 2M Karras / DPM++ SDE |
| Steps <br> (迭代步数) | 20 ~ 30 |
| CFG Scale <br> (提示词相关性) | 5.5 ~ 6.0 |
| Resolution <br> (推荐分辨率) | 512x768 / 768x1152 / 832x1216 |
✨ High-Res Fix Settings / 高清修复设置
- Upscale by (放大倍数): 1.5 ~ 2.0
- Hires Steps (重绘步数): 10
- Upscaler (放大算法): 4x-UltraSharp / R-ESRGAN 4x+
- Denoising Strength (重绘幅度): 0.3
- (建议保持较低到中低范围的重绘幅度，以避免风格过度覆盖并保持整体画面自然稳定 / Keep denoising in a low to moderate-low range to avoid excessive style override and preserve a natural, stable image.)