This LoRA is trained based on the official character art and CG art style of Dragon Soul Traveler.
The model strives to faithfully reproduce the game’s signature soft thick-painted anime aesthetic, delicate hair rendering, gem-like glowing eyes, and richly layered cinematic lighting.
Surprisingly, this model performs exceptionally well in complex dynamic poses and wide-angle perspective compositions. It greatly reduces anatomical distortion in high-motion actions such as jumping, running, and back views, while partially retaining the original game art style.
It supports various composition types, including facial close-ups, bust portraits, full-body character illustrations, and grand atmospheric scene creation.
recommend weight:0.6
本LoRA基于《龙魂旅人》官方立绘与cg画风训练而成。
模型尽可能地在还原该游戏标志性的柔和厚涂二次元质感、精致发丝表现、宝石质感发光眼瞳以及层次丰富的电影级光影。
意外的是，该模型在复杂动态姿态、广角透视构图场景中表现格外出色，能大幅改善跳跃、奔跑、背影等大幅度动作下的人体结构崩坏问题，同时部分保留原生游戏画风。
支持五官特写、半身人像、全身角色插画、宏大场景氛围创作等多种构图形式。
建议权重：0.6
Description
This v3.0 model was trained with a re‑optimized dataset based on the previous v1.0 and v2.0 versions.
v2.0 delivers steady style restoration and strong universal compatibility for most creation scenarios.
In comparison, v3.0 has a more distinctive stylized look, and adds a new adjustable feature: you can control the exposure level of character costumes（or level of sexy...） by tuning the LoRA weight.
The two versions have their own strengths, please choose according to your personal creative preferences.
recommend use at 0.6 weight
本次v3.0版本基于前序v1.0、v2.0版本，使用重新优化后的数据集完成训练。
v2.0版本画风稳定，泛用性更强，适配绝大多数创作场景。
与之相比，v3.0拥有更具辨识度的独特画风，同时新增了可控特性：可通过调节LoRA权重，自由控制角色服饰的暴露程度(或色情程度)。
两个版本各有优势，可根据自身创作需求灵活选择。
推荐权重0.6