Wan 2.2 VACE Fun 图像转视频

community/wan-2-2/vace-fun

Wan 2.2 VACE Fun 接收参考图像与文本提示词作为输入，利用 Wan 2.2 VACE Fun A14B 生成在身份一致性与动作表现力上高度协调的动画视频。

Prompt *

A dramatic, wide-angle tracking shot captures a lone figure in heavy tactical gear standing amidst the desolate ruins of a post-apocalyptic landscape during golden hour. The camera slowly dollies in from a low angle, gradually shifting focus from the vast, debris-strewn terrain to the figure, highlighting his resolute posture and intricate suit details. The sky is painted with streaks of clouds and contrails, bathed in warm orange light as the sun sets just above the horizon, creating a stunning contrast with the harsh, industrial structures in the background. The man wears a bulky yellow survival jacket over reinforced black tactical pants, strapped with various utility belts, pouches, and a compact backpack. His face is obscured by a high-tech gas mask equipped with communication devices and a small camera mounted on top. In the distance, a large, rusted rocket or industrial silo stands tall, its structure weathered and surrounded by thick plumes of dust rising from the ground, as if recently disturbed. The camera employs a shallow depth of field, bringing sharp clarity to the rugged textures of his jacket, the reflective surface of his helmet, and the scattered wreckage underfoot, while the background remains softly blurred. Slight handheld shake and subtle rack focus shifts emphasize the tension of the moment, evoking a feeling of isolation and determination as he surveys the wasteland for signs of life or dange

Image *

Resolution (W:H)

Number of Frames

Frames Per Second

Seed

Steps

Number of denoising iterations; more steps refine detail and stability but take longer.

Guidance Scale

Controls how strongly the output adheres to the prompt versus allowing creative variation.

Shift

Offsets the diffusion sampling schedule, trading stability for stronger motion/style as the value increases.

Idle

The rate is $0.10 per second, 1 second equals 16 frames.

Wan 2.2 VACE Fun 简介

Wan 2.2 VACE Fun 是一款基于提示词驱动的图像参考到视频生成工具，可将静态图像与文字描述转化为连贯的动画视频。该工具以 Wan 2.2 VACE Fun A14B 模型为核心，平衡了主体一致性、动作表现力和视觉保真度，是创意原型设计与故事叙述的理想选择。

Wan 2.2 VACE Fun 允许您通过提示词为参考图像赋予动态效果，生成既能保持人物特征又具风格统一的短视频。它专为需要电影化迭代、角色动画或快速可视化的艺术家、创作者和开发者设计。您只需提供输入，它便可输出富有表现力的 MP4 视频片段，方便预览与分享。

Wan 2.2 VACE Fun 的关键模型

Alibaba PAI Wan 2.2 VACE Fun A14B

该模型是整个流程的核心，负责将您的参考图像与提示词转化为可生成视频的潜在序列。它重点关注以主体为核心的动画生成与身份一致性保持。详细信息可参考其模型卡。

Wan 2.2 Text-to-Video A14B（HIGH/LOW）

这组文本到视频模型通过分阶段去噪实现速度与细节之间的最佳平衡。HIGH 版本注重纹理细节与画面保真度，而 LOW 版本则侧重运动稳定性。两种版本的权重均可通过 Comfy-Org/Wan_2.2_ComfyUI_Repackaged 获取。

Wan 2.2 VACE Fun 使用指南

必需输入

首先，设置 Prompt（提示词）并上传 Reference Image（参考图像）。您需要提供提示词以引导动作与艺术表达方向，并加载一张图像作为主体参考。

可选输入与控制项

高度（px）、宽度（px）、帧数（Number of Frames）和每秒帧数（FPS）用于控制视频输出的分辨率、长度和播放流畅度。您还可以通过 Sampling Steps 微调细节与稳定度，通过调整 Shift 平衡动作与纹理表现。Seed（随机种子）可设定为复现性结果，也可变更以生成不同版本。这些参数能让您在保持主体一致的前提下优化最终效果。