Veo 3.1 AI视频生成器
文本、图像与参考引导视频
使用Veo 3.1通过文本提示、首尾帧引导、参考图像、横竖构图以及Lite/Fast/Quality选项制作可控视频片段。
提示词 *
模式
参考图片 (1-2 张首尾帧图片)
拖拽图片到这里或 浏览
JPG、PNG、WEBP
宽高比
质量
种子 (可选)
Watermark (可选)
Auto-Translate Prompt
Translate non-English prompts before generation
准备生成视频
输入提示词并点击生成
Veo 3.1 model details
A deeper guide to Veo 3.1 input modes, quality choices, reference control, vertical video, and production settings in GemiOmni.
Veo 3.1 is for more controlled video work: text prompts, image-to-video, reference-to-video, vertical or horizontal framing, and quality modes. Use it when the clip needs a tighter relationship between prompt, reference images, camera movement, and final placement.
What do Text, Image, and Reference modes mean?
Text mode creates a scene from the prompt alone. Image mode uses first and last frame guidance for stronger visual control. Reference mode can use multiple images to guide subject, style, or composition when consistency matters more than pure exploration.
- Text-to-video for new scenes
- Image-to-video for frame control
- Reference-to-video for consistency
- Up to 3 reference images
How should I choose Lite, Fast, or Quality?
Use Lite or Fast when you want faster testing and prompt iteration. Use Quality when the shot needs better motion, light, detail, or production value. Reference-to-video currently uses Fast mode, so switch modes deliberately when you move between reference control and final-quality testing.
- Lite/Fast for iteration
- Quality for final candidates
- Reference mode uses Fast
- Test prompts before final runs
Which output settings does Veo 3.1 support?
Veo 3.1 in GemiOmni supports 16:9 and 9:16 output, Auto aspect where the selected image mode allows it, plus seed and watermark controls. It is especially useful when a single scene needs controlled framing for social video, ads, and cinematic placements.
- 16:9 and 9:16
- Auto for supported image modes
- Seed support
- Watermark control
Veo 3.1 vs Kling or Wan: when should I switch?
Choose Veo 3.1 for controlled cinematic scenes and reference-driven output. Try Kling for dynamic physical motion, Wan for stylized or flexible video prompts, Seedance for people and movement, and Hailuo for quick visual exploration.
- Veo 3.1 for control
- Kling for motion
- Wan for style
- Seedance for people and action
What prompt structure works best for Veo 3.1?
Use a compact production brief: subject, action, camera movement, lighting, location, mood, and intended platform. If you upload references, explain whether each image controls identity, style, background, or first/last frame.
- Subject plus action
- Camera and lighting
- Reference roles
- Platform-specific framing
Veo 3.1功能
GemiOmni工作区中可用的控制项。
三种输入模式
从文本开始，用首尾帧引导动作，或用参考图像增强一致性。
Lite、Fast、Quality
根据提示测试或最终候选片段的需求选择成本和质量路径。
画幅与Seed控制
在模式支持时使用16:9、9:16、Auto、seed和水印设置。
如何使用
3个简单步骤。
选择模式
根据需要的视觉控制程度，选择文生视频、图生视频或参考生视频。
配置
选择Lite、Fast或Quality，并在支持时设置画幅、seed、水印和参考图像。
生成
点击生成，任务完成后下载。
常见问题
常见问题。