Start building with Nano Banana 2 Lite and Gemini Omni Flash
Key Points
- Nano Banana 2 Lite: 4s latency, $0.034 per 1K image
- Gemini Omni Flash: $0.10/sec, 10s video, conversational editing
- Available via Google AI Studio, Gemini API and Gemini app
Summary
Google DeepMind released two models for fast, cost-efficient multimodal media pipelines: Nano Banana 2 Lite (gemini-3.1-flash-lite-image) for ultra-fast image generation and Gemini Omni Flash (gemini-omni-flash-preview) for high-quality video generation and conversational editing. Both are available in Google AI Studio, the Gemini API, and the Gemini Enterprise Agent Platform, with rollout to consumer surfaces (Gemini app, Search AI Mode, Google Flow, etc.).
Key Points
- Deployment and endpoints
- Available today in Google AI Studio, Gemini API and Gemini Enterprise Agent Platform; also rolling to consumer surfaces and Gemini app.
- Nano Banana 2 Lite (gemini-3.1-flash-lite-image)
- Target: rapid ideation and high-throughput pipelines; recommended upgrade from gemini-2.5-flash-image.
- Latency: ~4 seconds text-to-image.
- Cost: $0.034 per 1K-resolution image.
- Trade-offs: prioritizes speed and cost while retaining prompt adherence, character consistency and readable in-image text.
- Gemini Omni Flash (gemini-omni-flash-preview)
- Target: conversational video editing and multimodal video generation using text, image and short video inputs.
- Pricing: $0.10 per second of output (public preview), current generation limit: 10 seconds.
- Strengths: natural-language edits, multimodal referencing, text-action sync, and real-world knowledge for scene construction.
- Known limitations: no audio reference upload or scene extension via API yet; API accepts up to 3s video refs but they aren’t processed correctly; some character consistency and panning limitations.
- Integration tips
- Swap gemini-2.5-flash-image → gemini-3.1-flash-lite-image for immediate latency/cost gains.
- Chain models: generate images with Nano Banana 2 Lite, then feed them to Omni Flash to create animated clips.
- Use the Interactions API to preserve session history and support up to three sequential edits in multi-turn workflows.
- Safety and tooling
- Both models use SynthID watermarking and verification via the Gemini app, Chrome integration and Search.
Resources
- Google AI Studio playground, Gemini API docs, Nano Banana and Omni Flash prompting guides, and demo apps (Anywhere, Space Lift, Omni product studio) are available to get started.