Vidu 刚刚发布了全球首个【主体参照生成视频】新功能，太强了
赶紧上线体验一下了，以下是我的实际测试过程。
新特性如下：
角色类的单一主体图片，均能保持主体一致
功能：支持单一主体角色可控，多种类型角色可控（如真人、动物、奇幻角色等），支持多种画风角色可控（如写实、2D、3D等）
角色一致性
prompt：A woman with a delicate, elegant face, short brown hair, and bright red lips, dressed in a shimmering dress, smiles warmly at the camera
单一主体图片
2D角色一致性
prompt：The girl with black hair stood by the seashore, when suddenly she heard her name being called. She turned around, her eyes filled with confusion, scanning the horizon for the source of the voice
多种画风角色图片
选择主体
物体类的单一主体图片，均能保持主体一致
功能：支持任意单一物体的一致可控、单一物体的较高细节还原、商业广告场景的应用
物体一致性
prompt：A can of Harbin beer is placed in a vast desert, with bubbles rising from its surface, creating a refreshing visual. The scene features golden sand dunes under a bright blue sky, with the sun casting warm light that highlights the can's green and silver design. The camera uses a close-up shot to focus on the can and the bubbling beer, emphasizing the contrast between the cold beverage and the arid surroundings
单一物体图片
角色的面容、半身、全身的一致可控
功能：支持单一角色面容、半身、全身的一致可控
- 重点保持主体面容一致，需仅截取单一主体的清晰面容图
prompt：A young girl with large, expressive eyes and short black hair is dressed in a simple, elegant Chinese traditional robe. She sits quietly in an ancient palace, her delicate hands holding an open book as she reads with deep focus
单一主体的面容图
截取单一主体的清晰面容图
- 重点保持主体半身一致（面容 上半身服饰），需截取单一主体的清晰半身图
prompt：As the clouds dissipated, unveiling the serene visage of a princess attired in a scarlet robe and crowned with regal splendor, her beauty shone with a refreshing elegance, a blend of delicacy and poise
单一主体的半身图
截取单一主体的清晰半身图
- 重点保持主体全身一致（全身特征），需截取单一主体的清晰全身图
prompt：A cheerful cartoon snowman, wearing a crown and holding a Starbucks cup, with the Starbucks logo clearly visible, takes a sip of coffee. The cup features Starbucks' iconic green and white logo, which blends seamlessly with the Yeti's colorful and fun design. The background is still bright red, and the snowman's happy expression indicates that he is enjoying his coffee. The visual style remains consistent with the original, maintaining a bright and whimsical atmosphere
单一主体的全身图
截取单一主体的清晰全身图
实测下来，参考图片生成视频的效果很不错，不过距离商业片还是有一段距离，加油吧~
我们创建了自己的AI影视空间：
访问链接：https://xinzhiaigc.feishu.cn/wiki/Gn17wrpXliOzJzkL6WvcExlcnms?from=from_copylink
