Hey there,

At I/O last week, Google announced their new video model called Gemini Omni. This next generation release was briefly touched on last week in the recap of I/O but I wanted to cover it in greater depth today. So this week we’ll look at what makes Omni different from past video models, where you can use it, and the updates to Flow that came with this new model. 

What Gemini Omni actually is

Omni is Google’s new family of video models that can use any input type to create high-quality video outputs. But more than that, Omni has the reasoning capabilities that allow it to understand our world. These reasoning capabilities mean that Omni can understand physics, history, and science and can accurately use that information when generating videos. This comprehensive understanding is what makes this a “world model”. 

The Omni model can generate videos from text, image, video, and audio inputs which enables the model to build off your ideas no matter what format they’re in. These generated videos can include audio such as sound effects, dialogue, background noise, and more. All videos created with Gemini Omni can be easily identified by their SynthID watermark through tools like the Gemini app and Google Search. This makes it easy to identify content created by Omni and helps combat the flood of AI content on social media and the internet.

One of the most significant changes has to do with conversational editing. Users can easily edit their videos using natural language. Omni allows you to quickly and easily make changes to a video without needing to regenerate the entire thing. Effectively this means you can make small targeted changes to a video without losing the elements that are working properly. We’ve seen advancements like this in the image generation space as well with great results. 

How to access it

Omni Flash is available in a few places. The two main platforms are the Gemini app and Google Flow. You’ll need a paid Google plan to access Omni on both places but each offers a unique way to create with Omni. For more casual users, the Gemini app is a great place to explore Omni in an otherwise familiar place and style. For more intricate creation, Google Flow is the better option. Alternatively, you can access Omni for free via YouTube Shorts and the YouTube Create app. The Gemini team has also stated that Omni will be coming to the Gemini API in the coming weeks so you can develop your own apps soon.

What’s new in Flow

We’ve actually covered Google Flow in a past newsletter but the platform has seen some significant changes since then. 

• New Character Creation Screen: This is where you can create, edit, and manage any characters that you want to appear consistently within your Flow videos. By creating these characters beforehand, you are giving Omni something it can reference throughout the creation process. Ultimately this makes it easier to create polished, consistent characters. 

• Google Flow Agent: Enables a new collaborative approach to video generation and editing. This new agent can make recommendations for your project but also create multiple versions of your video to help give you more optionality. The Flow agent is also able to organize your creations into folders for easier storage and access. 

• Google Flow Tools: You can now create custom video tools within Flow. These can be anything from tools that let you draw within your videos, create animations, or even add text to your creations. Google also offers a number of premade tools you can use today. 

Google Flow will also be getting a new iOS and Android app soon which means you can create on the go without needing a laptop.  

Examples of Omni in Action

Now that we’ve covered what Omni is, let’s take a look at a few examples of what it can create. The first example showcases Omni’s ability to understand real-world knowledge to create a video of the Boston Tea Party without actually naming the event or location.

There are also some great examples of the physics being tested which allows you to see how the model handles motion and forces.