Loading...
Loading...
Google has introduced Gemini Omni, a native multimodal foundation model that processes and generates text, images, audio and video within a single architecture. Positioned to enable conversational editing and cross-modal reasoning—such as changing video scenes or characters with a simple prompt—Omni aims to replace separate generative pipelines with a unified developer surface. The first variant, Gemini Omni Flash, is available now in Google’s consumer apps (Gemini app, Google Flow, YouTube Shorts) for paid tiers, while enterprise API access via Vertex AI is promised in the coming weeks. The rollout signals a push to broaden multimodal capabilities for creative, marketing, and technical workflows once APIs become available.
A single native multimodal model simplifies development by replacing separate pipelines for text, image, audio and video, enabling unified APIs and new creative or automation workflows. Tech teams should plan for changes to content generation, editing tools, and infrastructure once enterprise access arrives.
Dossier last updated: 2026-05-21 10:23:06
Google DeepMind and Google announced Gemini Omni, a multimodal generative model family whose first release, Gemini Omni Flash, can take combined inputs (video, images, audio, text) and generate or edit high-quality videos via conversational prompts. Launched into the Gemini app, Google Flow and YouTube Shorts, Omni emphasizes iterative, context-preserving edits, consistent characters and improved physics-aware rendering, letting users transform scenes, alter action, change lighting, and refine shots across multiple turns. Google positions Omni as bridging photorealistic synthesis with reasoning grounded in world knowledge—history, science and culture—to produce more coherent, believable outputs. The rollout signals a push toward more integrated, multimodal creative tools across Google's consumer and creator products.
Google I/O 2026：Gemini Omni 世界模型——「模拟物理世界」的多模态 AI
Google announced Gemini Omni, a native multimodal foundation model that accepts and generates text, images, audio and video from a single architecture, with a focus on conversational video editing and coherent multimodal reasoning. The first model, Gemini Omni Flash, is live today in Google’s Gemini app for U.S. subscribers on paid tiers (AI Plus, AI Pro, AI Ultra), but the API for enterprise use via Vertex AI is promised only “in the coming weeks,” limiting immediate enterprise integration. Google positions Omni as collapsing separate generative pipelines into one system for cleaner developer surfaces and better cross-modal consistency, making it attractive for marketing, training, technical visuals and creative workflows once API access broadens.
At Google I/O 2026 Google announced Gemini Omni, a multimodal “omni” model claimed to be the most capable in the Gemini family, able to process and generate across text, images, video and audio. Demis Hassabis of Google DeepMind positioned Omni as supporting “any input to any output” and conversational editing — for example, changing video characters or backgrounds with a single sentence. Google also unveiled Gemini Omni Flash, the first Omni-family variant, available immediately in the Gemini app, Google Flow and YouTube Shorts, with API access promised later. The release signals expanded multimodal capabilities integrated into consumer and developer products.