内容来源：https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-omni-builders/

内容总结：

谷歌发布Gemini Omni，开发者已用AI视频生成玩出花
在今年的I/O大会上，谷歌推出了旗下全新Omni系列的首个模型——Gemini Omni Flash。该模型主打“对话式”视频创作与编辑，用户只需通过文字、图片、视频或音频参考，即可生成高质量视频，甚至还能直接编辑现有视频。由于Omni将直观的物理理解与Gemini的现实世界知识相结合，其生成的内容更加逼真自然。

目前，开发者已率先接入Omni，并基于该模型打造出多款令人眼前一亮的功能演示，以下是其中五个典型案例：

1. 自由切换机位与视角
Omni让视频编辑变得异常简单——用户可随时调整摄像机角度、切换环境背景，甚至实现电影级变焦，同时保持原始场景的连贯性。开发者Leon Lin就利用这一能力，从约20个不同机位拍摄了一位站在城市中央的女性：近景、远景、正面、侧面、俯拍、仰拍，背景也从人行道、斑马线到街道不断变换，行人、汽车、电车和各类建筑穿插其间，视觉效果极为丰富。

2. 改造你身边的世界
Omni允许用户随意改变视频中的剧情或物体，生成自己根本无法实拍的内容，同时确保画面逻辑自洽。开发者Carlos Santana仅凭语音指令，便将一段户外场景从白天切换为黑夜，添加阴云和雨声，让树叶变黄，地面覆盖积雪——整个环境瞬间“换季”。

3. 让日常物品“活”起来
在Google Flow中，Omni可将手绘草图转化为逼真视频，用户能通过涂鸦控制元素的运动方式，并决定它们是否出现在最终成品中。开发者Pan用这种玩法让日常物品变得趣味十足：柠檬变成潜水艇在海洋中起伏，咖啡杯里的浓缩咖啡化作热气球载人飞天，两只辣椒化身喷火龙，火柴变火箭，剪刀则成了觅食的鲨鱼。

4. 一键换装不同视觉风格
Omni支持用户为视频注入全新美学风格，只需输入参考图或直接用自然语言描述需求，模型便能融合生成风格统一的片段。开发者Jerrod Lew就尝试让一段女子街拍视频，在真人实景、动漫、黏土动画等多种风格间丝滑切换，整个过程丝毫不打断人物的自然行走。

5. 把创意构想变成现实视频
Hyperagent团队利用Omni为三个不同概念生成了视频：在空荡荡的公园画面中叠加绿化景观，呈现设计改造的前后对比；将数据拟人化为一位动画教授来讲解商业仪表盘；还将待办清单游戏化，用角色闯关的方式展示任务完成过程。

Omni将知识理解与创造力结合，让用户能够按照现实世界的逻辑构建场景。目前，用户可自行在Gemini应用、Google Flow、Google AI Studio、Gemini API或Gemini企业智能体平台中体验该功能。

中文翻译：

看看5位开发者用Gemini Omni做出了什么

在今年I/O大会上，我们推出了Gemini Omni Flash——Omni系列的首款模型。Omni让视频创作和编辑变得像对话一样简单。你可以通过文本、图像、视频或音频参考来生成高质量视频，甚至可以编辑自己的视频。由于Omni将对物理的直觉理解与Gemini的现实世界知识相结合，它生成的输出更加逼真。

我们最近向开发者开放了Omni的访问权限，而开发者们已经用它完成了各种项目。以下是一些引起我们注意的演示。

切换角度与视角。

Omni让视频编辑变得极为简单。你可以改变摄像机角度、切换环境、应用电影级变焦——全程不会丢失原始场景的叙事主线。

开发者Leon Lin（X平台账号@LexnLin）充分利用了这一能力，从大约20个不同角度拍摄了一位站在城市中央的女性。你能从一切可想象的角度看到她：近景和远景、正面和侧面、俯拍和仰拍。有些镜头拉近，有些保持静止。背景也在变化：她站在人行道、斑马线和城市街道上，周围是行人、汽车、电车和各种建筑。

改造周围的世界。

Omni让你能够改变视频中发生的事情，创造出你永远无法亲自拍摄的内容。你可以编辑动作或替换不同的物体，而Omni仍会维持一个连贯统一的场景。

以开发者Carlos Santana（X平台账号@DotCSV）的这段视频为例。他仅用语音就改变了一个户外场景：将光线从白天切换到夜晚，加上阴云密布的天空和雨声，把树叶变成橙色，并让地面覆盖积雪。

让日常物品动起来。

借助Google Flow中的Omni，你可以把草图变成逼真的视频，用涂鸦来引导各个元素的运动方式。你还可以决定这些手绘元素是否成为最终成品的一部分。

例如，开发者Pan（X平台账号@sebatheepan）用涂鸦趣味性地改造日常物品。一颗柠檬变成了在海洋中浮潜的潜艇。一杯浓缩咖啡变成了载着人飞越天空的热气球。两个辣椒变成了一条沉睡的龙，受惊时会喷出火焰。一根火柴被重塑为一艘火箭飞船。剪刀变成了一条寻找零食的可怕鲨鱼。

尝试不同风格或视觉效果。

好奇你的视频换成完全不同的美学风格会是什么样子？借助Omni，你可以改变风格、应用特效，创造出符合你设想的效果。通过输入参考来定义你想要的外观，或者直接用自然语言描述你的需求。Omni会将这些融合在一起，生成一段风格统一的视频。

开发者Jerrod Lew（X平台账号@jerrod_lew）在Google Flow中借助Omni测试了这一功能，用四种不同的动画风格渲染了一段女性走在街上的视频。原始片段流畅地在真人实拍、动漫、黏土动画等风格之间转换，全程不曾打断那位女性自然前行的步伐。

让创意活起来。

Hyperagent团队（X平台账号@hyperagentapp）利用Omni将三个不同概念可视化呈现为视频。他们将景观设计叠加到一段空旷公园的视频中，制作出改造前后的设计提案。他们将数据拟人化，用一位动画教授来解释业务仪表盘。他们还将待办事项游戏化，制作了一段角色逐项完成任务清单的视频。

Omni融合了知识与创造力，让你能够创造出遵循现实世界逻辑的场景。现在就试试吧——在Gemini应用、Google Flow、Google AI Studio、Gemini API或Gemini企业代理平台中均可使用。

英文来源：

See what 5 builders are making with Gemini Omni
At this year’s I/O, we introduced Gemini Omni Flash, our first model in the new Omni family. Omni makes creating and editing videos as easy as having a conversation. You can generate high-quality videos from text, image, video, or audio references, and even edit your own videos. And since Omni combines an intuitive understanding of physics with Gemini’s real-world knowledge, it creates more realistic outputs.
We recently gave developers access to Omni, and builders have already used it for a range of projects. Here are some of the demos that caught our eye.
Switch angles and perspectives.
Omni makes editing videos incredibly easy. You can change camera angles, switch environments, and apply cinematic zooms — all without losing the thread of your original scene.
Builder Leon Lin (@LexnLin on X) took full advantage of this capability, capturing a woman standing in the middle of a city from about 20 different perspectives. You see her from every angle imaginable: up close and far away, head on and in profile, from above, and from below. Some shots zoom in, while others hold still. And the background shifts: She stands on sidewalks, crosswalks, and city streets, surrounded by pedestrians, cars, trams, and a range of different buildings.
Transform the world around you.
Omni lets you change what’s happening in a video and create something you never could have filmed yourself. You can edit the action or swap in different objects, and Omni will still maintain a coherent, cohesive scene.
Take this video from builder Carlos Santana (@DotCSV on X), for example. He changes an outdoor scene using only his voice, switching the lighting from day to night, adding cloudy skies and rainy sounds, turning the leaves orange, and covering the ground in snow.
Animate everyday objects.
With Omni in Google Flow, you can turn sketches into realistic videos, using doodles to guide how individual elements move. And you can decide whether these hand-drawn elements become part of your final product.
Builder Pan (@sebatheepan on X), for instance, uses doodles to playfully transform everyday objects. A lemon becomes a submarine bobbing in the ocean. A cup of espresso is now a hot air balloon transporting someone through the sky. A pair of hot peppers form a sleeping dragon that breathes fire when startled. A match is recast as a rocket ship. And scissors turn into a scary shark looking for a snack.
Play with different styles or visual effects.
Curious what your video would look like with a completely different aesthetic? With Omni, you can change styles and apply effects to create something that matches your vision. Define the look you want by inputting references, or just describe what you’re looking for in natural language. Omni blends that together to create a cohesive clip.
Builder Jerrod Lew (@jerrod_lew on X) put this to the test with Omni in Google Flow, rendering a video of a woman walking down the street in four different animation styles. The original clip fluidly shifts from live-action to anime to claymation and more, never interrupting the woman’s natural forward progression.
Bring ideas to life.
The team at Hyperagent (@hyperagentapp on X) used Omni to visualize three different concepts with video. They layered landscaping into a video of an empty park to create a before-and-after design proposal. They personified data with an animated professor who explains business dashboards. And they gamified a to-do list with a video showing a character clearing tasks.
Omni blends knowledge with creativity, letting you create scenes that follow real-world logic. Try it in the Gemini app, Google Flow, Google AI Studio, the Gemini API, or the Gemini Enterprise Agent Platform.