Gemini Omni
FreemiumContent creators who shoot videos on their phones and want to reshape that footage without touching a timeline will find Gemini Omni a genuinely different kind of tool. It is not a video editor in the traditional sense — there are no layers, no keyframes, no trim handles. You describe what you want in plain language, and the model figures out how to apply that to your video. That is the core idea behind Gemini Omni, Google DeepMind's first model in the Omni family, launched at Google I/O on May 19, 2026.\n\nGemini Omni Flash — the first model released in this family — accepts text, images, audio, and existing video as input, and produces video output. The fact that it takes any combination of those inputs is what separates it from earlier tools. You are not limited to text-to-video prompting. You can hand it a reference image of a character, a sketch of a flying machine, and a clip of someone holding out their hand — and ask it to combine all three into a single cohesive output. That kind of multi-input referencing was not reliably possible before without significant manual compositing work.\n\nWhat makes the editing workflow distinctive is that it builds across turns. Each prompt you send continues from where the previous one left off, maintaining scene consistency. If you transport a violinist to a new environment in the first turn, then ask it to make the violin invisible, then ask it to shift the camera angle behind the performer's shoulder — each of those changes stacks without breaking the scene. In practice, this means you are not regenerating from scratch every time. You are refining. That iteration loop is what most creators will actually use it for.\n\nThe physics grounding is worth mentioning specifically. Gemini Omni draws on Gemini's broader world knowledge — including physics, biology, history, and scientific reasoning — to make generated footage feel more plausible. When you ask it to show a marble rolling down a chain-reaction track, it does not just move something in a straight line. It applies the kind of motion you would expect from gravity and surface friction. Whether you want fluid dynamics, light reflection off surfaces, or the natural behavior of soft materials, the model has been trained to treat those as real constraints rather than cosmetic suggestions.\n\nStyle transformation is another core use case. You can take a real video and convert it to voxel art, monochrome line drawing, holographic projection, stop-motion clay, or a puppet aesthetic — all through a natural language prompt. The original scene structure stays intact while the visual style changes. This has obvious applications for social media, short-form content, and any creator who needs to take one piece of footage and republish it in multiple visual formats without a design team behind them.\n\nReference-based editing gives you more precise creative control. You can pull in a reference image — say, a sketch or a photograph of an object — and ask Gemini Omni to insert that object into your video in a physically plausible way, complete with realistic lighting and integration with the scene. This means character swaps, object replacements, and environment changes can be driven by visual references rather than purely descriptive text. For filmmakers or content teams working with brand assets, this is practically useful in ways that purely text-driven generation is not.\n\nThe model also handles text-within-video more intelligently than earlier systems. It does not just overlay text — it understands how on-screen text should connect to what is happening in the frame. If you ask it to label items as they appear, or sync words to a beat, it treats that as a comprehension task rather than a formatting one.\n\nGemini Omni is accessible through the Gemini app, Google Flow (Google's AI creative studio built for filmmakers and video teams), YouTube Shorts, and YouTube Create. For subscribers to any Google AI paid plan — Plus, Pro, or Ultra — it is available today. Free users get access to Gemini Omni Flash on YouTube Shorts and YouTube Create at no cost. API access for developers and enterprise customers was announced as coming in the weeks after launch.\n\nAll content created with Gemini Omni in the Gemini app, Google Flow, or YouTube is marked with an imperceptible SynthID digital watermark, and C2PA Content Credentials are embedded as well. This means any video it produces carries verifiable metadata about how it was made, which is increasingly relevant as content verification becomes a real concern across publishing and social platforms.\n\nA higher-tier Omni Pro model has been confirmed by Google DeepMind, though no release date has been set. The team has indicated it will arrive when the performance improvement over Flash is substantial enough to justify a separate tier. For now, Flash is the only publicly available model in the Omni family.\n\nThe pricing structure ties into Google's existing subscription model. Free users get 50 daily Google Flow credits and access to a basic set of features including Gemini Omni Flash, Veo 3.1, and the ability to use Google Flow tools for usage-only tasks. Google AI Plus at $7.99 per month gives 200 monthly Google Flow credits and adds creation mode for Google Flow Tools, 1080p video upscaling, video-to-video editing, and higher image generation limits. Google AI Pro at $19.99 per month expands to 1,000 monthly credits and includes AI credit top-ups, plus access to the full Gemini app with 3.1 Pro, YouTube Premium Lite, and 5 TB of storage. Google AI Ultra at $99.99 per month gives 10,000 monthly credits, 4K image and video upscaling, the highest image generation limits, and the highest access tier for the AI agent. The $199.99 per month Ultra tier gives 25,000 monthly credits and everything in the $99.99 Ultra tier, alongside 30 TB of storage and YouTube Premium.\n\nFor solo video creators, marketers producing social content, educators building explainer videos, or filmmakers experimenting with style — Gemini Omni sits at a practical price-to-capability ratio that most standalone video generation tools have not matched. The freemium entry point via YouTube Shorts is particularly notable because it removes the need to sign up for anything to start testing the model's output quality.