Prince_Xiang_HunyuanVideo_1_5_lora

LoRA Model Card: svjack/Prince_Xiang_HunyuanVideo_1_5_lora

Character Wang Xiang: Artistic Portrait Synthesis

Fine-tuned T2V Adapter: hy_t2v_xiang_lora-step00002500-comfy.safetensors

Base Model: Compatible video synthesis foundation models: https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged

4 Step workflow: https://huggingface.co/svjack/Prince_Xiang_HunyuanVideo_1_5_lora/blob/main/hunyuan_video_1.5_720p_t2v_xiang_lora.json

Core Strengths:

• Character Consistency: Faithfully renders the distinct identity of Wang Xiang, maintaining consistent facial features and form across diverse scenarios.
• Aesthetic Detailing: Captures nuanced elements like skin texture, hair, fabric movement, and expressive gestures.
• Dynamic Poses: Effectively handles a variety of complex human poses with natural anatomical proportions.
• Artistic Style: Produces stylized portraits with a focus on atmospheric lighting and sensual elegance.

Inference Dataset: demonstrates inference samples of this lora model.

• https://huggingface.co/datasets/svjack/Xiang_hunyuan_1_5_t2v_frontal_videos
• https://huggingface.co/datasets/svjack/Xiang_hunyuan_1_5_t2v_different_style_videos

Optimized Text-to-Image Prompts for ComfyUI

Example 1: Gray Hoodie (Subtle Sophistication)

Prompt:

王翔戴着银色细框眼镜，镜片后深邃的眼眸泛着微光，灰色高领帽衫垂落的衣角随着他垂眸时的呼吸微微起伏，
露出锁骨处若隐若现的皮肤光泽。他指尖无意识地摩挲着帽衫褶皱，腕表带滑过手背时带起一阵若有似无的香气，
垂落的发丝扫过锁骨凹陷处，喉结在吞咽动作中泛起微红，
睫毛在眼下投出细碎阴影，仿佛随时会从镜片后溢出灼热的凝视。

Key Characteristics:

• Accessory Detailing: Glasses and watch are integrated naturally into the character's portrayal.
• Lighting Control: Subtle shadows and highlights add depth and mood.
• Texture Quality: Focus on fabric folds and the delicate rendering of skin and hair.

<video controls autoplay src="https://huggingface.co/svjack/Prince_Xiang_HunyuanVideo_1_5_lora/resolve/main/%E7%8E%8B%E7%BF%94%E5%B8%BD%E8%A1%AB(1080).mp4"></video>

---

Example 2: Gray Attire (Intellectual Ambiance)

Prompt:

王翔站在黄昏的书房窗前，逆光勾勒出他白嫩肌肤的流畅轮廓，深灰色丝绸衬衫半敞着领口，
下身穿灰色西裤，露出锁骨处若隐若现的淡青血管。他左手轻抚着窗边的古董书架，
右手无意识地撩起垂落的发丝，修长的指尖在书脊上留下蜿蜒的轨迹。镜头聚焦在他微蹙的眉峰与挑起的唇角，
睫毛在颧骨投下细碎阴影，腰肢随着转身动作向镜头倾斜，缎面衣料在转身时泛起丝绸般的光泽。
背景虚化成暖金色光斑，唯有他腰间若隐若现的银链腰带与书页翻动的沙沙声在画面中若即若离。

Key Characteristics:

• Lighting: Backlighting creates a dramatic silhouette and highlights skin and fabric textures.
• Composition: Intellectual setting with dynamic posing against a bookshelf.
• Details: Emphasis on subtle anatomical features and the interplay of light on silk.

<video controls autoplay src="https://huggingface.co/svjack/Prince_Xiang_HunyuanVideo_1_5_lora/resolve/main/%E7%8E%8B%E7%BF%94%E7%81%B0%E8%A1%A3%E6%9C%8D(1080).mp4"></video>

---

Example 3: By the Window (Casual Allure)

Prompt:

王翔下身穿蓝色牛仔裤站在落地窗前。他侧身倚着雕花窗棂，左手轻抚过垂落的发丝，
右手无名指在玻璃上划出蜿蜒水痕，指节泛着珍珠光泽。逆光勾勒出他脖颈处细腻的绒毛，
锁骨凹陷处泛着微光，喉结随着吞咽动作轻轻滚动。黑色半袖露出腰际一道若隐若现的红痕，
脚踝处缠绕的银链随着身体晃动折射出细碎光斑。他忽然转身，发丝扫过肩头时露出后颈处淡青的血管，
唇角微扬时睫毛在眼下投出颤动的阴影，指尖抚过窗棂雕花时，指节处的珍珠戒指在暗处泛着幽蓝微光。

Key Characteristics:

• Atmosphere: Intimate, casual setting with reflective surfaces creating depth.
• Composition: Focus on expressive gestures and the interaction with the window.
• Details: Attention to accessories (ring, chain) and the play of light on skin.

<video controls autoplay src="https://huggingface.co/svjack/Prince_Xiang_HunyuanVideo_1_5_lora/resolve/main/%E7%8E%8B%E7%BF%94%E7%AA%97%E5%89%8D(1080).mp4"></video>

---

Example 4: Blue Robe (Artistic Ambiance)

Prompt:

王翔赤裸的身躯笼罩在深蓝色丝绸睡袍下，睡袍边缘垂落的流苏随着他半蜷的姿势轻轻摇曳。
他侧身倚靠在雕花雕花红木沙发的曲线扶手上，左手无意识地抚过微微上扬的下巴，
右手则垂落在膝头，指尖在丝质布料上留下若隐若现的褶皱。逆光从他身后斜斜切过，
勾勒出线条流畅的身躯轮廓，细腻的肌肤在光影交错中泛着珍珠般的光泽，
锁骨处的阴影与喉结的轮廓形成微妙的视觉张力。他微微前倾的脖颈曲线与低垂的眼帘构成暧昧的三角关系，
睫毛在脸颊投下细碎的阴影，唇角若有似无的弧度暗示着某种未尽的邀约。
沙发背后的落地窗映出他半透明的睡袍，隐约可见腰际处的腰线与腿部线条，脚踝处的丝带缠绕着脚趾，
随着他轻微的呼吸起伏若隐若现。画面左上角的留白处，
一缕飘动的窗帘与窗外朦胧的月光形成虚实呼应，强化了他慵懒而危险的诱惑感。

Key Characteristics:

• Environmental Integration: Character is seamlessly blended into an opulent, atmospheric setting.
• Color and Light: Dramatic use of backlight and color (blue robe) to create mood and focus.
• Dynamic Elements: Subtle movements in the robe and body posture add a sense of life.

<video controls autoplay src="https://huggingface.co/svjack/Prince_Xiang_HunyuanVideo_1_5_lora/resolve/main/%E7%8E%8B%E7%BF%94%E8%93%9D%E8%A2%8D(1080).mp4"></video>

---

Example 5: Frontal Inference Dataset Grid

Inference Dataset: demonstrates inference samples of this lora model in indoor domain.

• https://huggingface.co/datasets/svjack/Xiang_hunyuan_1_5_t2v_frontal_videos

• Xiang Tease Grid Video <video controls autoplay src="https://huggingface.co/svjack/Prince_Xiang_HunyuanVideo_1_5_lora/resolve/main/%E7%8E%8B%E7%BF%949%E6%A0%BC(2048).mp4"></video>

---

Example 6: Different style Inference Dataset Grid

Inference Dataset: demonstrates inference samples of this lora model in different domain.

• https://huggingface.co/datasets/svjack/Xiang_hunyuan_1_5_t2v_different_style_videos

• Xiang Style Grid Video <video controls autoplay src="https://huggingface.co/svjack/Prince_Xiang_HunyuanVideo_1_5_lora/resolve/main/%E7%8E%8B%E7%BF%94style9%E6%A0%BC(2048).mp4"></video>

---

Technical Parameters

Usage Guidelines in ComfyUI

Workflow Integration

In ComfyUI, use the LoraLoader node to apply this adapter. Connect the MODEL and CLIP outputs from the base model loader to the LoraLoader, and then route the MODEL and CLIP outputs from the LoraLoader to the sampler and text encoder nodes, respectively. Multiple LoRAs can be applied sequentially.

Trigger Words

• Use "王翔" as the primary trigger for character consistency.
• Include descriptive terms related to the scene and mood for best results.

Weight Adjustment

• 0.9-1.1: Strong character presence, maintains distinctive features of Wang Xiang.
• >1.1: Experimental, may produce exaggerated characteristics.

License

CC-BY-NC-SA 4.0 (Non-commercial, share-alike) Community Hub: https://huggingface.co/svjack/Prince_Xiang_HunyuanVideo_1_5_lora/discussions

Training Methodology

This LoRA adapter was trained using low-rank adaptation techniques focusing on:

• Character Consistency: Maintaining stable identity and features across varied prompts. character has face:

• Aesthetic Quality: Enhancing the rendering of textures, lighting, and poses for artistic portraiture. The model demonstrates effective adaptation for generating consistent character portrayals within diverse narrative scenarios.