Google has launched Gemini Omni Flash, a video generation and editing model, in the Gemini app, Google Flow and YouTube Shorts. It is the first model in Google's new Omni family.
The system lets users combine text, images, video and some audio inputs to generate video clips, and supports conversational editing across multiple prompts. It is designed to preserve continuity in characters, scenes and motion as users refine a clip.
Omni builds on Google's broader Gemini strategy for multimodal artificial intelligence. It combines Gemini's reasoning with generative media tools, starting with video and later expanding to image and audio output.
Users can ask the model to alter existing footage, add objects or characters, change environments, or adjust style and camera angle through natural language prompts. Each instruction builds on the previous one, allowing iterative edits without restarting the process.
Gemini Omni Flash can also generate videos from mixed references, including text, images, video and voice inputs. At launch, audio input support is limited to voice references, with other audio inputs expected later.
Knowledge focus
Google positions the model as distinct from video tools that focus mainly on visual realism. Gemini Omni Flash draws on Gemini's broader knowledge base in areas such as history, science and cultural context to create clips that better match prompts requiring explanation or narrative structure.
The model also has a stronger grasp of physical behaviour, including gravity, kinetic energy and fluid dynamics. That is intended to improve how generated scenes handle motion and interactions between objects.
Another feature is the ability to create explanatory videos from short prompts. Google says the model can turn complex ideas into visual sequences, using Gemini's knowledge to connect language and imagery more coherently.
Avatar tools
Alongside video generation, Google is introducing avatar-based video creation using a person's own voice. The feature creates a digital version of the user so generated videos can resemble and sound like them.
Google is taking a more limited approach to broader speech and audio editing. It is still testing how to let users change audio and speech in videos responsibly before making that function more widely available.
The launch includes content authentication measures. Every video created with Omni includes Google's SynthID digital watermark, designed to be imperceptible, and can be verified through the Gemini app, Gemini in Chrome and Google Search.
Distribution plan
Gemini Omni Flash is being made available globally to Google AI Plus, Pro and Ultra subscribers through the Gemini app and Google Flow. The model is also rolling out at no cost to users of YouTube Shorts and the YouTube Create app.
The move extends Google's generative video tools beyond paid subscription products and into mainstream creator platforms. That could give it a broader test bed for consumer use cases in short-form video, where rapid editing and iterative prompting may appeal to creators working with existing footage.
Developers and business customers are next in line for access. Google plans to make Gemini Omni Flash available through application programming interfaces for developers and enterprise customers in the coming weeks.
The launch follows Google's earlier push into image generation and editing with Nano Banana, which it said had been used for tasks including restoring old photographs, working from sketches and visualising ideas. With Omni, Google is extending that multimodal approach into video creation from a wider range of inputs.