Google launches Gemini Omni for conversational multimodal video editing

TL;DR:

• Google has launched Gemini Omni, a multimodal generation family starting with Gemini Omni Flash, available in the Gemini app, Google Flow and YouTube Shorts; image and audio output modalities are described as forthcoming.
• The model accepts combined image, audio, video and text inputs and generates high-quality video grounded in Gemini’s “real-world knowledge,” with conversational, turn-by-turn editing presented as the headline interaction model.
• Google’s positioning leans on physics and continuity: characters stay consistent across edits, scenes “remember” prior turns, and the model is pitched as bridging “photorealism to meaningful storytelling” rather than as a raw text-to-video tool.

This is Google’s most credible frontier-multimodal push since Veo and a direct play at the YouTube creator pipeline. The conversational-editing framing is doing the strategic work: it positions Omni less as a competitor to OpenAI’s Sora or Anthropic’s adjacent work and more as a creator tool that lives inside Google Flow and YouTube Shorts. For UK creator-economy SMEs and ad agencies, the integration vector matters more than the raw model headline.

What the model actually does

Google describes turn-by-turn video editing through natural language: change specific elements, swap environments, redirect action, re-light, or change camera angle, with prior turns preserved. Demonstration prompts include making a sculpture out of bubbles, dimming room lights, ripple effects on touch, and transporting a violinist into a new scene. The Omni Flash variant is the entry point; Google flags the family will broaden across input/output modalities over time.

The “grounded in Gemini’s world knowledge” pitch sets up a capability contrast with raw generative video tools: Google is arguing Omni reasons about physics, history, science and cultural context to decide what should happen in a scene, not just render plausible pixels. Whether that holds in real creator workflows is the question Sora-class users will be testing this week.

Where this sits in the competitive picture

Frontier-video is now a multi-front competition: OpenAI’s Sora (and the rumoured Sora successor), Runway’s continuous releases, ByteDance’s video models inside TikTok, and now Gemini Omni inside YouTube Shorts. Google’s distribution advantage is YouTube Shorts itself – the largest creator funnel of any major AI lab – plus Google Flow as a professional surface. Anthropic, notably, has no consumer video product; that asymmetry is increasingly visible.

For UK SMEs, the operational read is: video-first creative work that was previously costed at hours per asset is now plausibly cost-able at minutes, but discoverability of the right model for the job is now an active workflow problem. YouTube Shorts integration is the most natural starting point for B2C UK brands; Google Flow is the more serious professional surface.

Looking forward

Watch how quickly Omni gets benchmarked by independent evaluators on the same physics-and-continuity claims Google leads with; the gap between launch demos and production fidelity is where frontier-video stories typically diverge. Expect a fast iteration cycle through 2026 as the Omni family adds image and audio output modalities. The Register podcast picking apart Google’s wider “AI enshittification” of search this week is a useful reminder that creator goodwill toward Google’s AI direction is not unconditional.