Google Omni and Flow · The Complete Guide to AI Video Creation and Editing.

Most people who saw Omni online think it is just another video generation model, comparing it to Kling or Seedance. That is a mistake. Omni is not a standard video model. It does not just take an image and animate it. It takes existing videos and changes them from the ground up, replacing styles, backgrounds, and clothing, adding visual effects, and creating a talking avatar based on your face. This is the Nano Banana of video. In session 23 of the AI Master Club, we reviewed Google's new model live, straight from the system. We saw a foosball table turn into a real field with players, a car in an office turn into a Ferrari on the road, a driver navigating a map of London, and a digital double climbing Mount Everest. All of this through a single prompt and a single click. Omni lives inside Flow, Google's new-old content creation studio. Flow is known to very few today, just like NotebookLM was a niche and weird tool when it launched, later becoming part of Gemini's memory. Flow is heading there. This guide is your entry point: what Omni can do, how to work with it correctly, and how to avoid burning your credits along the way.

What You Will Have by the End of This Guide

• Why Omni is not a standard video model, and the difference that changes everything
• How to log into Flow and start your first project, including the agent that manages your work
• Creating a digital double (avatar) in two minutes via your phone, and how it also appears in Gemini
• Visual effects on an existing video: taking a scene and turning it into anything
• Replacing clothing, backgrounds, and lighting inside a video with a single prompt
• Drone flyover and POV driving from a Google Maps map with an arrow
• 3D text that follows the body, and why it does not work in Hebrew
• The settings that save you credits: why Omni Flash instead of VEO, and how to work correctly
• All the tools, prompts, and questions that came up in the session

What is Flow and How to Access It

Digital Double · Creating an Avatar in Two Minutes

Important to know: avatar creation is locked for minors and requires identity verification via phone. As long as you work through Google tools, a watermark remains on the videos, intentionally, to indicate that the content was generated by AI. Do not rely on what you see on social media: Omni clones voices, faces, and movements to a degree that makes it hard to distinguish from the original.

Visual Effects · Taking a Scene and Turning It Into Anything

Visual Effect Prompt for an Existing Video

Take this video and keep everything exactly as it is: the person, the camera angle, the lighting, and the movement. Only change the following: after one second, transform [the object / area] into [the new thing you want]. For example: the table turns into a real football field with players running on it. Keep the transformation physically realistic and consistent with the rest of the scene. Do not change anything else in the video.

Replacing Clothing, Backgrounds, and Lighting · With a Single Prompt

Clothing and Background Replacement Prompt by Trigger

Take this video of me and keep my face, voice, and movements identical to the original. Use my finger snap as a trigger for edits: - On the first snap, change my outfit to [a formal suit]. - On the second snap, change the background to [a clean studio]. Keep the result realistic and natural. If a change does not land on the first try, prioritize a clean, believable result over speed.

Drone Flyover and POV Driving · From a Google Maps Map

Drone Flyover Prompt from a Map or Image

I'm attaching an image [a Google Maps screenshot / a building render] with an arrow marking the desired path. Create a smooth drone / aerial POV shot that flies along the marked path, from [start point] toward [end point]. Keep the geography and the buildings faithful to the source image. Make the motion cinematic and stable, like a real drone flyover.

3D Text and Pose Tracking

The AI Master Club opens the door for you: a live session every week, recordings, and a community. 47 ILS per month, cancel anytime.

Omni's limitations, to avoid wasting credits: Hebrew text inside the video does not work well. Scenes with many people and tiny details struggle. Almost nothing is perfect on the first try, and trial and error is required. Above all, Omni is not 100% faithful to the source, it clones voice, face, and movements, and sometimes invents text you did not say. Where it excels: replacing images, clothing, backgrounds, lighting, and visual effects.

Saving Credits · The Settings You Must Know

Flow is an entire amusement park, not just Omni. It is worth playing with the other tools: Storyboard Studio for developing plots and scenes, Character X-Ray for building a character with personality and clothing, Convert which takes several products and generates variations, Mockups to place a product on various mockups, Shot Explorer to get different angles from a single image, and Whisk, an entire image generation platform on its own. You have enough credits to experiment with everything.

All the Tools from the Session

The studio where Omni lives: video creation, avatars, effects, and editing. Accessed through Google Labs, working with a Gemini subscription

The avatar you created in Flow is also available here for quick video creation. The base subscription provides 1500 credits

Free, copyright-free source videos, great as raw material to feed into Omni and manipulate

For creating a map image with a route to feed into Omni for a drone flyover or POV drive

A source for fonts. You photograph a font and feed it to Omni when requesting 3D text on the video

Questions That Came Up in the Session

Yes, an active Gemini subscription is required. The good news: even the basic 20 dollar a month subscription gives 1500 credits, which is a lot. Creating a video in Omni Flash currently costs about 12 credits, so you can do dozens of trials. Without a subscription there are no credits, so you cannot work with it.

Every video comes out to about 8 to 10 seconds. You can extend it: inside Omni you connect multiple videos together (there is a plus button that adds another scene and continues from the last frame), or download the clips and connect them in an editing tool like CapCut while maintaining a consistent character.

The avatar is created inside Flow (in the Characters screen), and the creation runs through the Omni model. Once you have an avatar, you can also use it in Gemini for quick video creation, but it is more limited. For advanced work or at scale, it is better to do it directly in Flow.

Yes, from what we saw you can create multiple avatars. It also does full lip sync and knows how to speak in any language you request, including Hebrew, although the prompts themselves should be given in English.

Yes. A person is not required at all. You can feed in an image of a product or an object and use it: animate it, replace the background, change the lighting, or present it in various environments. If it is a person's image, you can replace the background and animate it to speak.

Probably not, and if so, not for long. Usually, a new Western model is open in the first two days and then they start reigning it in after complaints. For those who need to work with well-known figures, Chinese models (Kling, Seedance) are less restrictive, but they are also getting heat for it.

Not through Google tools. The watermark is intentional, to mark that the content was generated by AI and prevent fakes. It is possible that when an API opens it will be possible, but as long as you work through Flow and Gemini, the watermark stays.

And What is in the Next Session

The full recording of session 23, along with all prompts and links, is available to club members in the content library. Each session is saved with a written guide page, so you can scan in a minute and decide if it is for you. Try Omni, play around, and discover that the sky is the limit.

Frequently Asked Questions

Do you need a paid subscription to use Omni?

Yes, an active Gemini subscription is required. The good news: even the basic 20 dollar a month subscription gives 1500 credits, which is a lot. Creating a video in Omni Flash currently costs about 12 credits, so you can do dozens of trials. Without a subscription there are no credits, so you cannot work with it.

What is the maximum video length?

Every video comes out to about 8 to 10 seconds. You can extend it by connecting several videos together inside Omni, there is a plus button that adds another scene and continues from the last frame, or by downloading the clips and combining them in an editing tool like CapCut while maintaining a consistent character.

Is the avatar created in Omni, Flow, or Gemini?

The avatar is created inside Flow, in the Characters screen, and the generation runs through the Omni model. Once you have an avatar, you can also use it in Gemini for quick video creation, but that is more limited. For advanced work or at scale, it is better to work directly in Flow.

Can you create multiple avatars?

Yes, from what we have seen, you can create multiple avatars. It also does full lip-sync and knows how to speak in any language you request, including Hebrew, although you should give the prompts themselves in English.

Can you use just a photo of a person, without filming a video?

Yes. You do not need a person at all. You can upload a photo of a product or object and use it, animate it, replace the background, change the lighting, or present it in different environments. If it is a photo of a person, you can replace the background and animate it to speak.

Does it know how to create celebrities or famous people?

Probably not, and if it does, not for long. Usually, a new Western model is open for the first two days and then they start restricting it after complaints. For those who need to work with familiar figures, Chinese models like Kling and Seedance are less restrictive, but they are also getting heat for it.

Can you remove the watermark?

Not through Google tools. The stamp is intentional, to mark that the content was generated by AI and to prevent fakes. It is possible that when an API opens it will be possible, but as long as you work through Flow and Gemini, the stamp remains.