Google touts Gemini Omni trend: adding surprising twists to real footage
Google Flow's official account on June 3, 2026 promoted a new creative trend using its native multimodal video model "Gemini Omni" to add surprising twists to real-world footage. The company called the practice of "creating a surprising twist using real world footage" its current favorite trend and urged users to create their own (Google Flow).
Gemini Omni is a model formally unveiled by Google DeepMind at Google I/O 2026 on May 19, 2026. The first model, "Gemini Omni Flash," is positioned as able to "create anything from any input — starting with video." It takes any combination of text, images, audio and video as input and can generate and edit high-quality video. A key feature is its use of Gemini's world understanding of physics, biology, history and logic to achieve realistic motion and storytelling (Google Blog, DeepMind). Official demos show conversational editing of a violinist video, including changing the environment, hiding objects and altering the camera angle.
The model arrives as a replacement for the earlier video generation model Veo. Unlike Veo, it directly integrates Gemini's reasoning ability, applying a "world model"-like understanding of physics such as gravity and fluid dynamics to video generation. Against OpenAI's GPT-4o, Sora-family models and others such as Kling, Google has emphasized "any-to-any generation and editing in a single model," along with "integration of real-world knowledge" and "native editing" as differentiators (VentureBeat). To ensure transparency of AI-generated content, it supports the invisible SynthID watermark and C2PA metadata.
What matters is its aim to make video editing and generation iterable through natural-language conversation, eliminating the need for pro tools such as manual editing and switching between multiple apps. For creators, the dedicated studio Google Flow is provided, with integration with the image model Nano Banana also envisioned. Users can specify characters and styles with multiple reference images, with audio sync, video-to-video editing and character-ID consistency said to be possible, though detailed benchmarks such as resolution and length remain officially unpublished, with demo videos at the center.
The rollout began on Gemini apps and Google Flow the same day as the I/O announcement, with part of it offered free on YouTube Shorts that same week. Use requires a Google AI subscription (Plus, from $7.99/month; Pro, $19.99/month; Ultra, from $99.99/month), with looser limits on higher Ultra plans. Google Flow uses a credit system, with differences by tier such as 200 credits/month on Pro and 1,000–25,000 credits/month on Ultra (Google One). An API is said to be coming soon.
On X, the trend caught on across both official and user posts, with many videos sharing surprising twists added to real footage, such as "hybrid creatures," "adding a time traveler" and "LEGO-ization." Posts praising it ("Love this new trend!") and ones stating they were created with Gemini Omni Flash on @FlowbyGoogle lined up (Strength04_X). Ethan Mollick noted it is "fully multimodal" and able to edit video natively, with "excellent real-world logic" (emollick). On the other hand, opinions are divided, with concerns about the need for a subscription, the stability of generation quality, consistency in longer videos and feedback that it is "not yet perfect" (SidvexoAi). VentureBeat and Digital Trends reported on it in the context of the I/O announcement as an "any-to-any model" and a "replacement for your entire studio" (Digital Trends).