Start building with Nano Banana 2 Lite and Gemini Omni Flash
Summary
Google DeepMind released Nano Banana 2 Lite for fast, cheap image generation and opened Gemini Omni Flash to developers for video generation and conversational editing via the Gemini API and Google AI Studio.
Google DeepMind announced two model releases on June 30, 2026: Nano Banana 2 Lite, a speed-and-cost-optimized image generation model, and the developer preview of Gemini Omni Flash for video generation and editing. Both are available now through Google AI Studio, the Gemini API, and Gemini Enterprise Agent Platform.
What’s actually new
Nano Banana 2 Lite (gemini-3.1-flash-lite-image) is positioned as a drop-in upgrade for developers using the original Nano Banana (gemini-2.5-flash-image). Google quotes 4-second text-to-image latency and a cost of $0.034 per 1K-resolution image. The blog lays out the full Nano Banana family hierarchy: Lite for speed, Nano Banana 2 as the generalist, and Nano Banana Pro for complex professional tasks.
Gemini Omni Flash (gemini-omni-flash-preview) brings video generation and conversational video editing to the API for the first time. It accepts text, image, and video inputs and is priced at $0.10 per second of video output. Notable current limitations: output is capped at 10 seconds, video references up to 3 seconds are accepted by the API schema but not correctly processed by the model, and audio references and scene extension aren’t yet supported via the API.
The intended developer workflow is chaining the two: generate images quickly with Nano Banana 2 Lite, then pass them as references to Omni Flash for animation. Google mentions an Interactions API that maintains session history for up to three sequential edits in multi-turn experiences.
What it means for your config
These are API-level model identifiers, not local tooling configs — so there’s no .config file to update in the traditional sense. That said, if you have pipelines or agent configs that reference the old model string gemini-2.5-flash-image, Google explicitly recommends swapping it for gemini-3.1-flash-lite-image. Any environment variables, CI secrets, or orchestration configs that pin a Gemini image model ID should be reviewed.
For Omni Flash, the model identifier is gemini-omni-flash-preview — the preview suffix is worth noting. If you’re building production workflows around it, expect the identifier and behavior to change. The limitations around video reference processing and the 10-second cap mean you’ll want guard rails in any config or prompt template that feeds video inputs to this model.
The announcement doesn’t detail breaking changes in the API schema itself beyond the model swap, so if you’re already on the Gemini API, integration should be additive rather than disruptive. Check the developer docs (linked from the original post) for regional limitations and the full capability matrix before wiring anything into production.
Recommended next step
If you’re using the legacy Nano Banana model, the swap to Nano Banana 2 Lite looks straightforward — update the model ID and verify your outputs still meet quality expectations, since Lite explicitly trades some fidelity for speed. For Omni Flash, treat it as an exploration target: the preview label and the list of acknowledged limitations (especially the broken video-reference processing) suggest it’s not production-ready yet. Google provides demo apps you can remix to test the chained Nano Banana → Omni Flash workflow, which is worth a look before committing engineering time.
Read the full announcement on Google DeepMind Blog → Start building with Nano Banana 2 Lite and Gemini Omni Flash
More Google DeepMind Updates
Accelerating the frontiers of scientific discovery: Google's $40M commitment to the Genesis Mission
Google is committing $40M in AI tokens and cloud credits to the DOE's Genesis Mission, giving national lab researchers access to tools like AlphaEvolve, AlphaFold 3, AlphaGenome, WeatherNext, and AlphaEarth Foundations.
Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
Google DeepMind ships three new Gemini models targeting agentic workloads: 3.6 Flash (cheaper and more token-efficient than 3.5 Flash), 3.5 Flash-Lite (350 tokens/s throughput at rock-bottom pricing), and 3.5 Flash Cyber paired with the CodeMender security agent. Gemini 4 pre-training is also confirmed underway.
Our approach to bioresilience
Google DeepMind and Isomorphic Labs published a joint bioresilience strategy covering how they prevent misuse of AI models and deploy them for outbreak prevention, detection, and response. The post outlines more than 15 partnerships and several concrete technical initiatives.