Google will showcase Omni at Google I/O 2026 (May 19–20). Excited to see how this next-generation multimodal model advances AI-driven video creation and editing workflows. video by AIDRIVING #geminiomni
Gemini Omni is a cutting-edge multimodal video generation model developed by Google DeepMind. It enables video creation, editing, and remixing with flexible inputs such as text, images, video clips, and audio. With advanced scene consistency, camera control, and audio generation capabilities, Gemini Omni is suitable for advertising, content creation, and educational video production.
Gemini Omni processes multiple input formats to generate corresponding video content. For instance, when provided with an anime-style countryside sunset image, the model can produce a video that maintains the original composition, character design, and color palette, adding only subtle natural motion such as a gentle breeze moving the dress, hair, and sunflowers, along with drifting particles and slowly moving clouds. In another example, given a video clip of a person driving with accompanying text instructions, the model can replace the figure with a specified character while preserving vehicle motion and background environment.
Gemini Omni processes multiple input formats to generate corresponding video content. For instance, when provided with an anime-style countryside sunset image, the model can produce a video that maintains the original composition, character design, and color palette, adding only subtle natural motion such as a gentle breeze moving the dress, hair, and sunflowers, along with drifting particles and slowly moving clouds. In another example, given a video clip of a person driving with accompanying text instructions, the model can replace the figure with a specified character while preserving vehicle motion and background environment.
Try it nowGemini Omni integrates multiple input signals into unified creative instructions, allowing users to complete video generation and adjustments within a single workflow.
Gemini Omni accepts text, images, video clips, and audio as input references, interpreting them as interconnected creative directives. Users may describe concepts through text, define visual styles with images, suggest motion using video clips, and guide overall tone with audio. The model synthesizes these signals to generate video content that aligns relatively closely with user intent.
Users can modify existing video content through text descriptions without manually adjusting timelines or re-editing from scratch. For example, instructions such as "remove the specified logo from the frame" or "replace the spaghetti on both plates with creamy pumpkin soup while keeping everything else unchanged" enable the model to perform targeted modifications while preserving original composition, motion, and visual style.
Based on existing video clips, users can generate new versions through text instructions without rebuilding from the beginning. For example, combining a "person walking by the sea" clip with product footage can yield cinematic television commercial-style content that blends lifestyle presentation with polished product visuals.
The model supports precise adjustments to specific objects or details within a video rather than regenerating the entire scene. Users can request modifications to particular elements while maintaining original camera movement, frame composition, and visual style, improving iteration efficiency.
Compared to previous models, Gemini Omni demonstrates improvements in input flexibility, generation duration, scene consistency, and output quality.
Beyond text and image prompts, Gemini Omni supports video clips, audio, and templates as reference materials. Users can combine different input types within a single creative process without separating creative intent by format.
Generated video length is expected to reach approximately 15 to 30 seconds, with relatively smooth pacing and transitions. Regarding cross-frame consistency, the model shows enhanced ability to maintain character identity, scene details, and environmental elements, with improved object permanence and multi-character interaction stability compared to earlier versions.
The model supports relatively precise control over camera movement, framing, and pacing through text descriptions, and can achieve multi-angle transitions within a single scene. For example, it can shift from a frontal view to a side profile while maintaining consistent character appearance and environment.
Gemini Omni can generate scene audio matched to visual atmosphere, including character dialogue, ambient sound, and sound effects. In avatar generation, the model can maintain facial features and identity consistency based on reference images, with lip synchronization and facial expression changes aligned to voice content.
The model applies to multiple fields requiring rapid video generation or adjustment, helping users with varying backgrounds reduce video production barriers.
Suitable for advertising prototype creation, pre-visualization, and commercial short film production. Creators can quickly generate proof-of-concept videos through text, adjusting camera language and visual style across multiple iterations to assist pre-production decision-making.
Applicable to short-form video and channel content creation. The model supports multi-segment video generation with consistent characters and visual styles, facilitating coherent series content creation, while generated audio can accommodate dialogue requirements.
Usable for product demonstration videos and brand content production. Through natural language descriptions, users can adjust product presentation, scene atmosphere, and visual tone within the frame, shortening the cycle from creative conception to final output.
Suitable for explanatory videos, operation demonstrations, and teaching content production. The model shows improved capability in maintaining text and formula logic, capable of generating footage including blackboard derivations and step-by-step demonstrations. Multi-angle camera switching also helps display specific operational details.
Follow Gemini Omni on Twitter to see the latest community creations, feature updates, and real-world video stories.
Google will showcase Omni at Google I/O 2026 (May 19–20). Excited to see how this next-generation multimodal model advances AI-driven video creation and editing workflows. video by AIDRIVING #geminiomni
Holllllyyyyyyyy @GeminiApp cooked 😳😳 🚨 Gemini Omni: New video model Here is the first output and see the text coherence , if this is not nano banana moment of video then what is ?? direct link for those who believes otherwise in comments
GOOGLE 🔥: An upcoming Gemini Omni video model from Google is expected to be much more advanced in video editing, capable of completing tasks like removing watermarks, replacing objects in the video, and more. It is also likely that Google will release 2 versions of this model,
🫨Google is creating a new Omni model with good video editing. Veo4? The original is on the left. Edited right. The new model also does a good job of removing watermarks from videos.
Sample video and early feedback 👀 > I won’t lie, this is one of the best video models I have seen, maybe not *the* best, but a really strong performance. I was particularly impressed by the prompt adherence (except for the one shot with the missing centerpiece), the model
