Consistent character
The person from the reference images — same face, hair, and outfit — walks through a neon-lit city street at night, cinematic tracking shot, shallow depth of field.
Free Reference to Video AI Generator
Upload one or more reference images and generate new videos that keep your character, object, and style consistent across every frame.
加载中
什么是参考图生视频 AI？
参考图生视频使用上传的图片，引导新生成视频中角色、产品或场景的外观。添加提示词描述动作和镜头运动；一致性因模型而异，可能需要多次尝试。
将角色与物体融入您喜爱的场景
在保持环境一致的同时，无缝添加新角色和物体，让您喜爱的场景更加丰富生动。借助参考生视频工具，只需点击几次即可将可爱的动物引入美丽花园，或在温馨咖啡馆的桌上添一杯热气腾腾的咖啡。
提示词
在巴黎圣母院前，三位女士正在自拍。
为您的主角/物体生成全新情境
使用参考图片引导角色或产品出现在新场景中。在提示词中描述新背景和动作，然后检查结果中主体外观的变化。
提示词
戴草帽的女士骑着自行车穿过玫瑰园。
将多张图片混合为单个连贯视频
我们的参考生视频工具可将多张图片无缝转换为流畅连贯的视频。只需上传您的素材——例如野餐、孩子玩耍与日落——AI 视频生成器就会在数秒内将其串联成一段优美故事。
提示词
在公园里，一个小男孩和一只小狗正在湖边的草地上野餐玩耍。
释放创意视觉的无限可能
借助参考生视频生成器，您可以轻松为社交媒体创作引人入胜的视频、在各种场景中展示产品，或制作打动人心、极具趣味性的视频作品。
流畅过渡与自然动作
参考生视频生成器会分析上传图片中的每个选定元素，不仅为人物和物体，还为整个场景生成平滑过渡且动作自然的动态视频！
满足各类创作者的参考生视频生成器
为视频剪辑创建重复出现的角色
只需上传包含不同主体和物体的图片，它们就会无缝融合进一致的视觉环境中，无需逐帧对齐即可构建平滑剪辑与连续角色。
如何将参考图片转换为视频？
只需 3 个简单步骤，即可创建高品质角色一致性 AI 视频
- 01
上传参考图片
上传一张或多张代表目标人物、物体或场景的参考图片。
- 02
设定一致性目标
选择要在整个视频中保持一致的元素，并输入动作与镜头提示词。
- 03
生成连贯视频
让 AI 施展魔法，生成视觉连贯、动作流畅的动态视频。
参考图生视频与其他视频工作流有何不同？
| 工作流 | 主要输入 | 最适合 |
|---|---|---|
| 文生视频 | 文字提示词 | 根据描述创建场景。 |
| 图生视频 | 起始图片和动作提示词 | 为提供的图片添加动作。 |
| 参考图生视频 | 参考图片和场景提示词 | 引导主体在新场景中的外观。 |
参考生视频提示词示例
Product hero shot
The product from the reference image rotates slowly on a marble pedestal, soft studio lighting, label and silhouette unchanged, premium commercial look.
Style-locked scene
A new scene in the same illustration style as the reference: the character waves at the camera in a sunny park, consistent colors and line work, gentle camera push-in.
常见问题
What is reference to video?
Reference to video is an AI video workflow where you upload one or more reference images — a character, a product, or a scene — and the model generates a brand-new video that keeps that subject's identity and style consistent in every frame. Instead of animating the exact picture you upload, the model treats your references as an identity guide while your prompt describes what should happen in the new video.
How many reference images do I need?
One clear reference image is enough to lock a subject. Adding more images — different angles, outfits, or lighting of the same subject — gives the model a stronger identity signal. The workspace accepts up to 4 reference images on models that support multiple references; the exact limit depends on the selected video model.
Can I keep a character consistent across videos?
Yes. Reuse the same reference images across generations and keep the character description in your prompt stable. Because the references anchor the identity, you can place the same character into different scenes, actions, and camera moves while the face, outfit, and proportions stay recognizable from video to video.
What's the difference between reference to video and image to video?
Image to video animates the exact picture you upload — your image becomes the first frame and the model adds motion to it. Reference to video does not reuse your picture as a frame at all: it generates a completely new scene and only carries over the identity and style of the subject in your references. Use image to video when you want this photo to move; use reference to video when you want this character or product to appear in new videos.
Is the reference to video generator free?
Yes. You can upload reference images, write a prompt, and generate a video for free in the browser — the workspace runs a full-quality generation without asking for a credit card. Heavier usage is covered by affordable credit plans, but trying the reference to video generator costs nothing.
Do I need to sign up?
No sign-up is required to try it. The generator works for guest visitors directly on this page; creating an account only becomes useful when you want to keep your generation history, favorites, and credits in one place across devices.
立即免费开始制作惊艳、一致的角色视频
来源和页面信息
Google Cloud 的视频提示词指南涵盖主体、动作、场景和镜头运动。该指南介绍 Veo 的提示词；FreeCreate 模型提供的控制选项和结果可能有所不同。
“Specificity helps avoid generic outputs.”Google Cloud 视频生成提示词指南维护者 FreeCreate ·
