视频模型 · 免费 · 无需注册

Wan
提示词生成器。

Wan 是 Alibaba 的开源视频模型,能遵循详细的长描述:人物、场景、运动和画面质感——光线、景别、机位角度和色彩。

它还支持负面提示词,所以生成器会把你的排除项输出为单独一栏,包括闪烁、面部扭曲等运动瑕疵。

示例输出夜间自拍短片
Horizontal 16:9 nighttime glamour selfie video. An original fictional adult woman in her mid-20s sits in a dark upscale evening interior at night. She has very long deep chocolate brown hair, smooth glossy blowout with loose waves, warm medium tan skin, warm brown eyes, glossy nude-rose lips. She wears a rich wine red deep plunging sleeveless evening dress and a delicate short tennis-chain necklace. She slowly tilts her head so her hair falls over one shoulder, brushes a strand behind her ear, then looks straight into the lens with a confident, faint smile. Throughout: natural blinking and breathing, her hair shifting gently with each movement. Camera: phone held at arm's length, slight natural sway, tight upper-body framing, slightly low. Low-key warm ambient lighting, key light from front-left and slightly below, approximately 2500-3200K warm amber, near-black background. Extreme photorealism, luxury social-media portrait, deep burgundy, warm tan and dark chocolate brown palette, sharpest on eyes and face, slow and calm pacing.
159 个词aspect_ratio: 16:9duration: 5 seconds+ 负面提示词

如何写出更好的 Wan 提示词。

01

写得越细越好

Wan 会用上这些额外信息;描述画面质感要和描述动作一样用心。

02

用好负面提示词栏

运动瑕疵和真实感问题,最好在负面提示词里处理。

03

本地也能用

提示词是纯文本,在 ComfyUI 或任何其他 Wan 界面中都能用。

Wan 示例。

全部示例 →

Wan 常见问题。

本地部署的 Wan 能用吗?+

能用。把提示词和负面提示词原样粘贴进你的工作流即可。

画幅比例和时长怎么设?+

在“画幅”和“运动”分区里设置;它们会作为单独的参数列出。