Google is pushing Gemini beyond conversation and image generation, this time toward a more direct kind of media creation. With Gemini Omni, the company is positioning AI to turn text, audio, images, video clips, and even mixed inputs into new video content automatically.
That shift matters because it moves the competition in generative AI toward a faster workflow for creators, businesses, and everyday users. Instead of stopping at chat responses or still images, Google is now focusing on video production and editing as the next major AI frontier.
A broader multimodal system with video at the center
Gemini Omni builds on the Gemini platform, which already handled text, images, audio, and video together. Google is taking that multimodal foundation further by making video generation and editing the main use case.
In Google’s framing, the system is designed to create “anything from any input,” according to CEO Sundar Pichai. That ambition shows that Google wants Gemini to do more than answer prompts; it aims to understand real-world context and turn it into digital content.
At the first stage, Gemini Omni is centered on generating videos from text, images, audio, and short video segments. The system then processes those elements into a new video without requiring long manual editing steps.
Editing through simple text instructions
Gemini Omni is not only a generator. It also introduces a more practical editing workflow, where users can change images or videos with simple text commands instead of professional editing software.
Google describes this as the next step in combining Gemini’s intelligence with its media rendering capabilities. The result is intended to produce more complex and realistic visual output than earlier generations of AI tools.
One of the demonstrations showed Gemini Omni creating a clay-style animation about protein folding. The output appeared as a stop-motion video with automatic voice narration matched to the topic.
That demo highlighted how generative AI is moving beyond simple visuals. It can now assemble educational and informational presentations that look more structured and purposeful.
Digital avatars and identity protection
The system also includes support for personal digital avatars. These avatars let users create virtual versions of themselves that can be used to generate videos automatically.
The idea is similar to OpenAI’s Cameos feature on Sora, but Google pairs the feature with identity safeguards. Users must complete identity verification before creating a digital avatar.
That process requires recording the face while speaking a set of numbers provided by the system. Google also applies SynthID digital watermarking to all videos made with Gemini Omni.
Those watermarks are meant to help people identify whether a video was produced by Google’s AI system. The company is placing that protection alongside the avatar feature to reduce the risk of misuse and deepfake abuse.
Gemini Omni Flash starts rolling out
Google introduced the first model in the lineup under the name Gemini Omni Flash. It is now available through the Gemini app, YouTube Shorts, and Google’s AI creative platform, Flow.
At launch, Gemini Omni Flash can generate videos up to 10 seconds long. Google says that limit is not a technical constraint, but a way to keep access open so more users can try the feature without heavy computing queues.
Support for longer videos is still being developed. In the meantime, the model is already being used in a range of creative scenarios, from someone receiving an award to a trip to the moon.
Google also showed editing examples, including removing an unwanted person from the background of a vacation video. DeepMind research engineer Gabe Barth-Maron described the experience as a new form of personalized meme creation.
Even so, the final quality still depends heavily on the prompt. More detailed instructions produce more accurate results, while vague prompts can cause the system to change elements that users may have wanted to keep.
