V4.0听说KREA2模型更加细致效果更好就炼了一个，确实很强，但是训练成本也是真的高，建议强度0.8-1I heard the KREA2 model offers finer detail and better results, so I trained one; it is indeed very powerful, though the training cost is truly high. I recommend a strength setting of 0.8–1.V3.9使用了新的训练方式，大幅度减少了画面的噪点，质感也会更好Version 3.9 uses a new training method, which significantly reduces image noise and improves image quality.在单独使用V3.9时，建议强度1When using V3.9 alone, a lora strength of 1 is recommended.V3.5版本更改了训练参数，优化了皮肤质感，减少了服装布料褶皱过多的问题Version 3.5 adjusted training parameters, optimized skin texture, and reduced the issue of excessive fabric wrinkles.建议强度0.8V3版本是我目前调整最满意的版本，优化了颜色饱和度和光影，同时优化了服装和背景的质感Version V3 is the one I am most satisfied with so far; I have optimized the color saturation and lighting, as well as the textures of the clothing and background.Recommended strength: 0.7–1; consider lowering the strength when using in combination with other LoRAs.触发词Trigger words<pre><code>@semi-realisticV3</code></pre>V2这个版本是巨大的优化改善了皮肤颜色，服装质感，手脚崩坏等问题。并且大幅改善了脸部的表现力Version V2 brings massive improvements.It addresses issues such as skin tones, clothing textures, and distortions in the hands and feet, while also significantly enhancing facial expressiveness.此LORA会产生动漫风格的脸和偏向写实风格的身体和服装（包括背景）建议强度0.7-1触发词<pre><code>@Semi-Realistic Animation</code></pre>This LoRA generates anime-style faces combined with semi-realistic bodies, clothing, and backgrounds.Recommended weight: 0.7-1Trigger word:<pre><code>@Semi-Realistic Animation</code></pre>

Semi-realistic animation style

听说KREA2模型更加细致效果更好就炼了一个，确实很强，但是训练成本也是真的高，建议强度0.8-1I heard the KREA2 model offers finer detail and better results, so I trained one; it is indeed very powerful, though the training cost is truly high. I recommend a strength setting of 0.8–1.

semi-realisticV4_000001200.safetensors

v4.0(krea2)

Files with SHA256: d0fce8241fb7cd0f302f7cd6e26327c814ed5bcd737bb377cf52490f08183909