“Black hole” or a new era of video generation?

Google has introduced a new family of generative models called Gemini Omni, and this update is already being described as one of the company’s most radical steps toward a complete rethinking of digital content production. This is not just another neural network for video generation, but an attempt to unify text, images, audio, and video into a single system capable of creating entire virtual worlds based on user prompts. The first product in the lineup is Gemini Omni Flash, available through the Gemini app and Google Flow service. Its core task is generating videos from any type of input data: text, photos, audio clips, or existing video. Unlike previous generation models, the system does not operate as a set of separate tools, but as a unified multimodal “engine” that simultaneously understands visuals, sound, and semantic meaning.

Google is effectively moving from content generation to reality modeling.

Inside the company, it is emphasized that Gemini Omni combines multimodal capabilities with an understanding of fundamental physical laws of the real world. This means the model does not simply “draw images,” but attempts to reproduce the logic of motion, light behavior, fluid dynamics, and object interactions. According to developers, the system draws on the broader Gemini knowledge ecosystem, allowing it to incorporate scientific and historical context when generating scenes.

One of the most striking features is video editing via natural language. Users can not only generate clips but also modify existing videos in a dialogue format: changing character actions, scenes, atmosphere, or event sequences while preserving narrative consistency. In effect, editing becomes a conversation with the system.

This is where the real market shift begins.

Where video editors, animators, and designers once worked through complex tools, editing software, and production pipelines, a large part of that process is now being transferred into text prompts. The user describes the result — the system constructs the scene.

Dumitru Erhan, Senior Director of Research at Google DeepMind, notes that at this stage the model can generate videos with sound up to 10 seconds long, although the company is already working on extending this limit. Even in this form, the technology demonstrates a level of detail that was recently considered experimental.

Koray Kavukcuoglu, CTO of Google DeepMind and Chief AI Architect at Google, emphasizes that the new system has a much deeper “understanding of the world” than previous models. This refers to the ability to account for physical processes and cause-and-effect relationships rather than just visual patterns.

One of the most discussed features of Gemini Omni is the generation of digital avatars. Users will be able to create characters that resemble themselves and voice them using their own voice. This continues the personalization trend that has already driven adoption in previous Google generative tools. Product lead Nicole Brichtova notes that such features have already proven highly popular in earlier systems.

Safety remains a key concern. Google has introduced restrictions on altering other people’s speech in videos to reduce abuse risks. All generated videos are automatically marked with an invisible SynthID watermark, allowing verification of content provenance and distinction between real and synthetic media.

The company also plans to expand capabilities in the future, including audio and static image generation within a unified model.

However, the main impact of this technology is not technical but economic and labor-related.

Gemini Omni brings the industry closer to a reality where video creation is no longer a craft but a dialogue with a system. Editing, color grading, animation, sound design, and even scripting logic are partially absorbed into a single interface — the text prompt.

This directly affects professions that for years were considered stable within the creative industry. Designers, video editors, motion designers, post-production specialists, and even parts of directing are gradually moving into the automation zone.

Where the industry once focused on accelerating workflows, it is now moving toward full abstraction. Instead of tools — a model. Instead of skills — prompt formulation.

Supporters argue this represents the democratization of creativity: anyone will be able to produce visual scenes without professional skills. Critics highlight the opposite side — the devaluation of professions built on years of experience and manual craftsmanship.

These changes are happening especially fast in short-form video and advertising content, where production speed matters more than depth. Here, generative models are already beginning to compete not with individual specialists, but with entire production studios.

In a broader sense, Gemini Omni represents another step toward what can be called the “synthetic content industry.” A space where a significant share of visual and audio material is produced not directly by humans, but by models trained on vast datasets of real-world content.

The irony is that the very designers, editors, and artists who taught the digital industry its visual language are now watching that language being reproduced automatically.

And if the trend continues, the question will no longer be whether AI can replace creative professions, but what role humans will retain in a process where the “image” itself becomes the output of a single line of text.