Google Launches Nano Banana 2 Lite and Gemini Omni Flash: 4-Second Image, 10-Second Video, Lightweight Creative Model Combo

On June 30, 2026, Google DeepMind quietly released two lightweight AI creative models: the image model Nano Banana 2 Lite (codename gemini-3.1-flash-lite-image) and the video model Gemini Omni Flash. The former generates 1K resolution images in about 4 seconds at a cost as low as $0.034 per image; the latter supports conversational video editing with an output cost of $0.10 per second. Both models can be chained via the Interactions API to create an end-to-end "text → image → video" pipeline. Google also open-sourced three demo applications (Anywhere, Space Lift, Omni Product Studio) showcasing potential in travel, interior design, and e-commerce scenarios.

Core Model Capabilities

Nano Banana 2 Lite: Fastest and Cheapest Image Model

• Speed: Generates a 1024×1024 image in about 4 seconds, one-fifth of Nano Banana 2 (20 seconds).
• Cost: $0.034 per image, roughly half of Nano Banana 2 and a quarter of Nano Banana Pro.
• Performance: Achieved Elo scores of 1255 (report 1) or 1251 (report 2) on Arena.ai, ranking fifth, outperforming the original Nano Banana Pro.
• Capabilities: Maintains prompt adherence, character consistency, and text clarity in images. Google recommends original users upgrade directly.

Gemini Omni Flash: Conversational Video Editing Model

• Input: Supports mixed text, image, and video inputs; outputs up to 10-second videos.
• Editing: Allows up to three consecutive rounds of editing via natural language, retaining context.
• Knowledge: Built-in Gemini world knowledge, can leverage common sense from history, biology, etc.
• Limitations: Does not yet support audio references or scene extension; video reference processing under 3 seconds is imperfect; limited character consistency during scene transitions.

Pricing and Competitor Comparison

Chinese vendors like ByteDance Jimeng and Kuaishou Kling price a 5-second video at about $0.4, roughly $0.08 per second, slightly lower than Omni Flash.

Chained Workflow and Demo Applications

Through the Interactions API, users can first quickly generate an image with Nano Banana 2 Lite, then use it as a reference input for Omni Flash to generate a video, and continue editing with natural language. Google released three open-source demos:

• Anywhere: Upload a selfie, Lite composites the portrait into a landmark scene, Omni Flash turns it into a dynamic video.
• Space Lift: Upload a room photo, Lite generates multiple renovation plans, Omni Flash creates a spatial walkthrough video.
• Omni Product Studio: A product white-background image is transformed by Lite into a contextual product shot, then Omni Flash converts it into an e-commerce ad video.

These features are integrated into Gemini App, Google Flow, YouTube Shorts, and other products, available for free.

Community Reaction and Industry Impact

Positive feedback focuses on cost and efficiency: Google Developer Relations team member Paige Bailey noted NB2 Lite has become the default image generation tool; enterprises like WPP, Figma, and Adobe have already integrated. Negative feedback includes: queue times over 30 seconds during peak hours, Chinese text rendering errors, occasional six-finger issues, and unstable artistic style transfer. Some developers anticipate the flagship model Gemini 3.5 Pro, originally scheduled for June release but reportedly delayed to July, with Google declining to comment.

Analysts believe Google's move is not a "rescue" but a parallel product strategy: flagship models address capability ceilings, while lightweight models tackle speed, cost, and workflow integration needs. As the quality gap among leading models narrows in mid-2026, models that embed into user workflows first may gain a commercial advantage.