Nano Banana 2 Lite & Gemini Omni Flash: Image to Video

Google shipped two creative-AI models on June 30, 2026 that are more interesting together than apart. Nano Banana 2 Lite is the fastest, cheapest model in Google’s Nano Banana image family — a text-to-image in roughly four seconds at $0.034 per 1K-resolution image. Gemini Omni Flash, launching in public preview, generates and conversationally edits video. The headline is not either model on its own; it is the pipeline they form — generate an image cheaply, then animate it — at a price and latency that make high-volume creative production practical.

For a marketing team that is the whole point. The expensive, slow part of creative has always been volume: a hundred product variations, a dozen ad cuts, localized versions for every market. A four-second image and a per-second video price change what is worth producing at all. Google’s own launch demos lean directly into this — one, “Omni Product Studio,” turns static product shots into cinematic e-commerce videos.

This guide covers what each model actually is, where Nano Banana 2 Lite sits in the four-tier family, what Omni Flash can and cannot do yet (it is a preview, with real limitations), how to chain the two, and what the release means for a marketing or e-commerce team that produces creative at scale. Every latency and price figure here is Google-stated.

Both models landed on June 30, 2026 through the Gemini API, Google AI Studio, and the Gemini Enterprise Agent Platform, with consumer surfaces following. Nano Banana 2 Lite is generally rolling out; Gemini Omni Flash is a public preview (gemini-omni-flash-preview) and available to developers for the first time, alongside placements in the Gemini app and Google Flow.

The naming is worth untangling, because Google now ships four image models under the Nano Banana banner. Nano Banana 2 Lite is the new, speed-focused entry; Nano Banana 2 is the generalist; Nano Banana Pro is the high-end option; and the original Nano Banana — the model we covered in our first Nano Banana guide — is now the legacy tier Google recommends upgrading off. Omni Flash sits outside that family entirely: it is a video model, not an image model.

Read in sequence, this is Google doing to video roughly what Nano Banana did to image generation a year ago — making a capable model fast and cheap enough that the constraint stops being the model and starts being your imagination. The person who leads Google AI Studio and the Gemini API put it more plainly the morning it shipped.

Nano Banana 2 Lite (gemini-3.1-flash-lite-image) is positioned as the fastest, most cost-efficient model in the family. Google’s two headline numbers tell the story: a text-to-image in about four seconds, at $0.034 per 1K-resolution image. That latency is the part that changes behavior — four seconds is fast enough to put image generation inside an interactive loop, a high-velocity pipeline, or a consumer feature where a user is waiting on the result.

The interesting claim is what Google says it did not trade away for that speed. Lite is described as keeping reliable prompt adherence, strong character consistency, and legible in-image text rendering — the three things that usually break first when an image model is optimized for throughput. In-image text in particular has been a perennial weak spot for generative image models; keeping it legible at this price point is the meaningful part of the claim, if it holds up in your own testing.

Where Lite fits is clearest against the rest of the lineup. Google now frames the Nano Banana family as a four-tier ladder — Lite for speed, Nano Banana 2 as the generalist workhorse, Nano Banana Pro for complex professional work, and the original Nano Banana as a legacy model to migrate off. Only Lite has published latency and price figures in the announcement; the table below uses Google’s positioning for the others rather than inventing numbers.

Gemini Omni Flash is the more experimental half of the release: a multimodal model that takes text, image, and video inputs and specializes in video generation and conversational editing. Google calls it a “high quality, cost-efficient model for video generation and conversational editing,” priced at $0.10 per second of video output. It generates 10-second clips today, with longer durations described as coming soon.

“Conversational editing” is the differentiator. Rather than re-prompting from scratch, you refine a clip in natural language — change the camera move, swap an element, adjust the action — and the model draws on Gemini’s broader world knowledge (history, biology, narrative logic) to keep edits coherent. It also supports multimodal referencing, combining image, text, and video inputs to steer a generation.

The two models are designed to hand off to each other. The intended pattern: generate a still cheaply and quickly with Nano Banana 2 Lite, then pass that image to Omni Flash to animate it into a clip — and keep refining, maintaining context across up to three sequential edits through the Interactions API. That image-to-video loop is the engine behind all three of Google’s launch demos.

The three-edit context window is the practical detail that makes this a workflow rather than a party trick. Because the Interactions API holds state across sequential edits, you can iterate on a clip conversationally — generate, adjust the framing, then tune the action — without losing the thread. For a fuller picture of where Omni Flash sits against the dedicated video models, our Omni vs Sora vs Veo 3 comparison covers the trade-offs, and it shares a lineage with the native computer-use work in Gemini 3.5 Flash.

Strip away the demos and the value to a marketing team is about volume economics. Most creative work that does not get made is blocked on cost and time, not ideas: the fortieth product video, the localized ad cut for a small market, the weekly batch of social variations. A four-second image and a ten-cent-per-second video move those from “not worth it” to “default.” That Nano Banana 2 Lite is rolling out inside Google Ads is the clearest signal of where Google expects it to be used.

The honest caveat sits alongside the opportunity. Generative media at this volume needs a review layer, not just a generate button: brand consistency, factual accuracy on product detail, and usage rights all still require a human in the loop — and Omni Flash’s preview limitations make that doubly true for video. The teams that win here will treat these models as a throughput multiplier on a governed creative process, not a replacement for one. That is the same balance we cover in our look at image-generation API pricing and the higher-end Nano Banana Pro for marketing.

The practical opening is narrow on purpose. Pick a single high-volume creative task you already do by hand — product-image variations, social cutdowns, localized banners — and rebuild just that one as a Nano Banana 2 Lite pipeline, with a human approving the output. Prove the quality and the cost on a real workflow before you add Omni Flash video on top, since the image side is generally available and the video side is still a preview.

From there the sequencing mirrors how we roll any generative system out with clients: prove value on one governed workflow, add the video step once the image quality earns trust, and only then widen to more creative types — keeping brand review and rights checks in the loop throughout. That scoping is exactly where our AI content engine and e-commerce engagements begin, before any tool commitment.