<span style="background-color:rgb(255, 255, 255);color:rgb(0, 0, 0);font-family:&quot;Color Emoji&quot;, system-ui, &quot;PingFang SC&quot;, PingFangSC-Regular, &quot;Microsoft YaHei&quot;, &quot;Hiragino Sans GB&quot;, &quot;Heiti SC&quot;, &quot;WenQuanYi Micro Hei&quot;, Arial, Helvetica, sans-serif, &quot;Apple Braille&quot;, &quot;Color Emoji Fix&quot;;font-size:14px">## V2.0 · Strong Style Edition A Gufeng / Xianxia style LoRA that enhances figure quality, posing, garment material and fabric detail, and lighting depth. Strong style transfer intensity. If you encounter broken hands, fused fingers amplified after loading the LoRA, or overly dark output, raise CFG to 2.0–4.0. No trigger word required. Control intensity via weight: 0.2–0.3 subtle texture ｜ 0.5–0.6 standard ｜ 0.7–0.8 strong style ｜ above 0.8 enters "mystery lottery" mode. Sample images are shown in order from 1.0 down to 0.3. Style characteristics - Reinforces Gufeng / Xianxia themes: costumes, poses, texture and overall stylistic consistency are noticeably improved - Enhanced lighting contrast, stronger depth of field, more solid and dimensional rendering - Heavier texture and spatial depth compared to the base model's native output Known trade-offs - To gain stronger style and tonal depth, part of Qwen-Image 2.1's native overall brightness is sacrificed - If you want to preserve the base model's clarity and prefer balanced output, use the V3.0 version instead Base model: Qwen-Image 2.1 ｜ Recommended weight: 0.3–1.0<span style="background-color:rgb(204, 235, 255);color:rgb(0, 0, 0);font-family:&quot;Color Emoji&quot;, system-ui, &quot;PingFang SC&quot;, PingFangSC-Regular, &quot;Microsoft YaHei&quot;, &quot;Hiragino Sans GB&quot;, &quot;Heiti SC&quot;, &quot;WenQuanYi Micro Hei&quot;, Arial, Helvetica, sans-serif, &quot;Apple Braille&quot;, &quot;Color Emoji Fix&quot;;font-size:14px">## V2.0 · 强风格版（Strong） 古风 / 仙侠风格 LoRA，主要融入强化了人物质感、姿态，服饰材质款式质感，光影层次。风格迁移强度较高。如果出现底模崩手、融手被加载lora后放大，或者画面很暗，请将CFG拉高到2.0~4.0 本 LoRA 无需触发词。加载后通过权重控制强度：0.2–0.3轻微质感 ｜ 0.5–0.6 标准 ｜ 0.7–0.8 强风格 | 超过 0.8 开启玄学抽奖模式。示例图依次为 1.0 至 0.3 风格特征 - 古风、仙侠题材强化：人物服饰、姿态、质感、整体画风统一性明显提升 - 光影对比增强，景深层次更突出，画面真实立体感更强 - 质感与空间层次较底模原生表现更厚重 已知取舍 - 为换取强风格与影调层次，会部分牺牲 Qwen-Image 2.1 原生的通体明亮感 - 若需保留底模通透感、追求均衡表现，请使用 V3.0 版本 底模：Qwen-Image 2.1 ｜ 推荐权重：0.3-1.0

Chinese Fantasy Xianxia Style LoRA | 仙侠古风

## V3.0 · Balanced Edition (Recommended) Expanded training set for this version. Strikes a balance between style reinforcement and the base model's native clarity and brightness — the recommended version for everyday use. If you encounter fused or broken hands, raise CFG to 2.0–4.0. Image quality and specific art style are determined by your prompt. The aesthetic core injected by this LoRA stays consistent. Style characteristics - Fully preserves Qwen-Image 2.1's native clarity and brightness — no gray cast, no crushed shadows - Adds stronger aesthetic layering: figure posing (graceful, dynamic), garment texture, unified art style, clearer spatial depth and depth of field - Covers Gufeng portraiture, Xianxia characters, and character photography Differences from V2.0 - V2.0 Strong: heavier style and tonal depth, partially sacrifices the base model's brightness and clarity - V3.0 Balanced: retains clarity while adding aesthetic layering (Recommended) - For extreme stylization / epic fantasy feel, use V2.0 Usage No trigger word required. Control style intensity via weight: - 0.2–0.3 subtle texture enhancement - 0.4–0.6 standard - 0.7–0.8 strong style - 0.9–1.0 extreme stylization — "mystery lottery" mode Base model: Qwen-Image 2.1 ｜ Not face-fixed, stackable with other character LoRAs ## V3.0 · 均衡版（推荐） 此次增加了训练集，在风格强化与底模原生明朗通透感之间取得平衡，为日常使用首选版本。如果出现基座融手、崩手，请拉高cfg值，例如2.0~4.0.画面质感和画风由提示词决定，lora注入的美学内核会持续保持。 风格特征 - 完整保留 Qwen-Image 2.1 原有的通透明朗感，画面不发灰、不压暗 - 融入更强的美学层次：人物姿态（柔美、张力等）、服饰质感，画风统一、空间纵深与景深更明确 - 题材覆盖古风人像、仙侠角色，与人物写真 与 V2.0 的区别 - V2.0 Strong：风格强烈、影调层次重，会部分牺牲底模明朗通透感 - V3.0 均衡版：保留通透感的同时兼顾美学层次（推荐） - 若追求极限风格化 / 奇幻史诗感，请使用 V2.0 使用方式 本 LoRA 无需触发词，加载后通过权重控制风格强度： · 0.2~0.3 轻微质感加成 · 0.4~0.6 标准 · 0.7~0.8 强风格 · 0.9~1.0 风格化极限区 —— 玄学抽奖模式 底模：Qwen-Image 2.1 ｜ 非固定人脸，可与其他角色 LoRA 叠加

Files with SHA256: a80f68caa1f02a2bbddb617c2e490bfc3c37f0ad1052a3d725ee36d8f0e81e46