<h1 id="true-first-person-female-body-pov">True First-Person Female Body POV</h1><p>这是一个专注于<strong>真实女性第一人称身体视角</strong>的 LoRA。</p><p>相机位于成年女性角色的眼睛位置，向下能够看到属于视角主体自己的手臂、身体、衣物、腿和脚，而不是第三人称俯拍、自拍或普通的“POV风格”构图。</p><h3 id="v2.0">V2.0 更新说明</h3><p>V2.0 进一步强化真实的第一人称身体视角，改善双手、手臂、躯干和双腿从画面下方自然延伸的效果。提升行走、坐姿、日常动作、人物互动及复杂场景的稳定性，并减少第三人称跟拍、后脑勺、漂浮肢体和身体缺失等问题。</p><p>核心提示词：</p><pre><code>true first-person body POV</code></pre><p>按需添加：</p><pre><code>adult female viewpoint
adult male viewpoint</code></pre><p></p><h2 id="recommended-settings">Recommended Settings</h2><ul><li><p>LoRA weight：建议从 <code>0.7–2.0</code></p></li><li><p>CFG、采样器和步数：根据底模调整</p></li><li><p>支持竖图与横图</p></li><li><p>竖图更适合表现完整身体、腿部和脚部</p></li><li><p>横图更适合双人互动、多人场景和环境构图</p></li></ul><p>如果第一人称效果不够明显，可以增加：</p><pre><code>looking down at her own body, her own hands visible, her own torso visible, her own legs visible</code></pre><p>如果出现第三人称镜头，可以尝试加入负面提示词：</p><pre><code>third-person view, external camera, overhead camera, selfie, mirror selfie, full character seen from outside, detached hands, extra arms, extra legs</code></pre><h2 id="example-prompts">Example Prompts</h2><h3>日常场景</h3><pre><code>true first-person body POV, adult female viewpoint, looking down at her own clothed body, cardigan, jeans, hands resting on her lap, sitting at a cafe, warm window light, photorealistic</code></pre><h3>户外场景</h3><pre><code>true first-person body POV, adult female viewpoint, walking on a forest trail, her own hands, jacket, trousers and hiking boots visible, mountains in the distance, natural daylight, photorealistic</code></pre><h3>双人互动</h3><pre><code>true first-person body POV, adult female viewpoint, holding hands with an adult partner, her own arm, clothes, legs and shoes visible, walking through a city at night, affectionate atmosphere, photorealistic</code></pre><h3>多人场景</h3><pre><code>true first-person body POV, adult female viewpoint, sitting with a diverse group of adult friends, her own hands and lap visible, friends interacting naturally, rooftop lounge, cinematic lighting, photorealistic</code></pre><h2 id="training-dataset">Training Dataset</h2><p>本 LoRA 使用了 80 组图文配对进行训练，训练集包含：</p><ul><li><p>单人、双人和多人构图</p></li><li><p>站立、坐姿、行走、运动和日常动作</p></li><li><p>拥抱、依偎、牵手、慢舞等亲密但非露骨的互动</p></li><li><p>咖啡馆、家庭、户外、旅行、工作坊和社交活动</p></li><li><p>酒吧、爵士酒廊、夜店、赌场、屋顶酒廊和晚宴等成人社交场所</p></li><li><p>38 张竖图和 42 张横图</p></li><li><p>多种分辨率与画面比例</p></li><li><p>东亚、南亚、黑人、拉丁裔、中东、混血及多族裔成年人</p></li></ul><p>训练数据均为<strong>明确成年人、完整着装、非露骨内容</strong>。</p><h2 id="notes">Notes</h2><ul><li><p>LoRA 的核心能力是第一人称身体构图，不是固定人物或固定服装。</p></li><li><p>人物族裔、肤色、体型、服装和场景可以通过提示词单独控制。</p></li><li><p>复杂的双手动作、镜面反射和多人肢体接触仍可能受到底模能力影响。</p></li><li><p>权重过高可能产生夸张的向下视角、重复肢体或身体比例问题。</p></li><li><p>本模型没有使用裸体或性行为图片训练，因此不能保证 NSFW 构图的稳定性。</p></li><li><p>请仅用于明确的成年角色。</p></li></ul><hr /><h2 id="english-description">English Description</h2><h1 id="true-first-person-female-body-pov">True First-Person Female Body POV</h1><p>This LoRA is designed to create a <strong>genuine first-person body perspective from the viewpoint of an adult woman</strong>.</p><p>The camera is positioned at the character’s eye level, looking down at her own arms, hands, clothing, torso, legs, or feet. It is intended to avoid third-person overhead shots, selfies, and images that merely resemble a POV composition.</p><h3 id="trigger-words">Trigger Words</h3><pre><code>true first-person body POV, adult female viewpoint</code></pre><p>Place the trigger phrase near the beginning of your prompt.</p><h3 id="suggested-weight">Suggested Weight</h3><pre><code>0.7–1.0</code></pre><p>Start lower and increase the weight if the first-person body perspective is not strong enough.</p><p>Helpful supporting phrases:</p><pre><code>looking down at her own body, her own hands visible, her own torso visible, her own legs visible</code></pre><p>Suggested negative prompt:</p><pre><code>third-person view, external camera, overhead camera, selfie, mirror selfie, full character seen from outside, detached hands, extra arms, extra legs</code></pre><h3 id="dataset">Dataset</h3><p>The LoRA was trained on 80 image-caption pairs featuring:</p><ul><li><p>Solo, couple and group compositions</p></li><li><p>Sitting, standing, walking, sports and everyday activities</p></li><li><p>Hand-holding, embraces, cuddling and slow dancing</p></li><li><p>Indoor, outdoor, travel and social environments</p></li><li><p>Adult nightlife venues such as cocktail bars, lounges, clubs, casinos and evening events</p></li><li><p>38 portrait and 42 landscape images</p></li><li><p>Multiple resolutions and aspect ratios</p></li><li><p>Diverse adult subjects, including East Asian, South Asian, Black, Latina, Middle Eastern and mixed-race representation</p></li></ul><p>All training subjects are clearly adults. The training dataset is fully clothed and non-explicit.</p><h3 id="important-notes">Important Notes</h3><p>This LoRA focuses on perspective rather than a specific character, ethnicity, body type, outfit or location. These details should be controlled through your prompt.</p><p>Very complex hand interactions, mirrors and crowded scenes may still depend heavily on the capabilities of the base model. Excessively high LoRA strength may produce distorted limbs or an exaggerated downward camera angle.</p>

Files with SHA256: b8d935ba1ae28077340ce70caf0adc1336f75ee170534b49dc2dfe50677cee94