Google has released two new AI models designed to make creative content generation faster and more affordable: Nano Banana 2 Lite for rapid image creation and Gemini Omni Flash for multimodal video editing. The image model generates pictures in approximately four seconds, while the video tool enables developers to create and edit 10-second clips using text, images, and natural language instructions.

How Fast Are These New Models Compared to Previous Versions?

Nano Banana 2 Lite represents a significant speed improvement over its predecessors. The model generates a 1K resolution image in roughly four seconds, compared to approximately 20 seconds for Nano Banana 2 and seven seconds for the original Nano Banana. In real-world testing, the model produced images in an average of three seconds, versus 19 seconds for Nano Banana 2. This dramatic acceleration makes the tool particularly useful for design reviews, rapid prototyping, and brainstorming sessions where quick iteration matters more than perfect polish.

Despite the speed gains, Google did not sacrifice quality entirely. On Arena.ai, a platform where humans compare and rate AI outputs, Nano Banana 2 Lite scored 1,251 for image generation, just 19 points below Nano Banana 2's 1,270 but 100 points higher than the original Nano Banana's 1,151. For image editing tasks, the new model scored 1,308, placing it 13 points above the original model, though slightly below Nano Banana 2's 1,387. This positioning reflects Google's strategy: prioritizing a balance between quality, speed, and cost rather than chasing maximum quality.

What Makes These Models Affordable for Developers?

Pricing is where Nano Banana 2 Lite truly differentiates itself. A single 1K resolution image costs approximately 3.9 cents using Nano Banana 2 and 6.7 cents with the original Nano Banana, but just over three cents with Nano Banana 2 Lite. For API usage, the model charges 25 cents per one million input tokens and $1.50 per one million output tokens. Gemini Omni Flash, the companion video model, costs 10 cents per second of video output, matching the pricing of Google's other speed-focused video generation model, Veo 3.1 Fast.

These cost reductions matter significantly for high-volume workflows. Developers running design systems, e-commerce platforms, or content creation pipelines can now generate images at scale without prohibitive expenses. Google emphasized that the models are designed specifically for scenarios where "speed and cost are critical".

Google

How to Integrate These Models Into Your Workflow

• Image Generation: Access Nano Banana 2 Lite through Google AI Studio, the Gemini API (Application Programming Interface), or the Gemini Enterprise Agent Platform for immediate use in development environments.
• Consumer Product Rollout: The model is rolling out sequentially across Google Search's AI Mode, the Gemini app, NotebookLM, Google Photos, Stitch, Google Flow, and Google Ads, bringing fast image generation to millions of end users.
• Video Creation Workflow: Combine Nano Banana 2 Lite with Gemini Omni Flash by generating images first, then animating them into videos using natural language editing instructions for seamless creative workflows.
• Interactive Editing: Use Gemini Omni Flash's natural language editing to modify videos with written instructions, maintain scene consistency by referencing images and text, and synchronize on-screen text with video movements.

Google demonstrated this integrated workflow by showing how images generated with Nano Banana 2 Lite can be passed directly to Gemini Omni Flash and converted into video. The company also released demonstration applications showcasing practical use cases, including virtual travel experiences, interior design visualization, and e-commerce video creation.

What Are the Current Limitations of Gemini Omni Flash?

While Gemini Omni Flash opens new possibilities for video creation, the current public preview version has notable constraints. The model can generate videos up to 10 seconds long, though Google stated it plans to support longer outputs in the future. The Gemini API does not yet support uploading audio references or extending scenes. Additionally, while the API accepts video references up to three seconds long, the model may not process them correctly. Character consistency in videos involving scene changes and camera movements remains an area for improvement.

Despite these limitations, Google positioned Gemini Omni Flash as a tool for developers and creators to experiment with AI-assisted video production without requiring advanced editing skills. The company is actively developing these features, indicating that future updates will expand capabilities.

How Does Google Address AI-Generated Content Transparency?

Both models incorporate SynthID, Google's invisible digital watermarking technology that identifies content as AI-generated. As the volume of AI-generated images increases with faster, cheaper generation, Google emphasized the importance of maintaining identifiability. The company also stated that it filters, labels, and red-teams training data to suppress harmful outputs, demonstrating commitment to responsible AI development.

Google's rollout strategy reflects its broader ambition to embed these models across both developer platforms and consumer-facing products. Nano Banana 2 Lite is available immediately through Google AI Studio, the Gemini API, and the Gemini Enterprise Agent Platform, with sequential expansion to Search, Gemini apps, NotebookLM, Google Photos, Stitch, Google Flow, and Google Ads. Gemini Omni Flash is currently available through Google AI Studio, the Gemini API, the Gemini app, and Google Flow in public preview.

The launch of these models signals Google's continued focus on making advanced AI capabilities accessible to developers and consumers alike. By prioritizing speed and cost efficiency alongside quality, Google is positioning itself to serve high-volume creative workflows where previous models proved too slow or expensive. As these tools roll out more widely, they may reshape how teams approach rapid prototyping, design iteration, and content creation at scale.