Google DeepMind Launches Nano Banana 2 Lite and Gemini Omni Flash

Google DeepMind released two generative media models on June 30, 2026, pushing into an increasingly crowded market for AI-powered image and video creation. The two models — Nano Banana 2 Lite, a speed-optimized image generator, and Gemini Omni Flash, a conversational video generation and editing model — are now available to developers through Google AI Studio and the Gemini API. For students and early-career builders, the most relevant detail is this: Google AI Studio is free to access, meaning anyone with a Google account can start experimenting without spending a dollar.

What Google Actually Released

Nano Banana 2 Lite (model ID: gemini-3.1-flash-lite-image) is the newest and leanest member of Google’s Nano Banana image model family. It generates text-to-image outputs in roughly 4 seconds and is priced at $0.034 per 1,000-resolution image — positioning it as the go-to option for high-volume workflows where per-unit cost matters. Google describes it as the recommended upgrade path from the original Nano Banana (gemini-2.5-flash-image), which is now considered the legacy tier. The model retains solid prompt adherence, character consistency, and readable in-image text despite its emphasis on speed.

The Nano Banana family now has four distinct tiers: the new Nano Banana 2 Lite for speed; Nano Banana 2 as the generalist workhorse; Nano Banana Pro for complex, accuracy-heavy professional tasks; and the original Nano Banana, which Google is actively steering developers away from. Beyond developer tools, Nano Banana 2 Lite is also rolling out to consumer surfaces including AI Mode in Search, NotebookLM, Google Photos, Stitch, Google Flow and Google Ads.

Gemini Omni Flash (model ID: gemini-omni-flash-preview) is a different kind of model entirely. First shown at Google I/O in May 2026, it reached the Gemini API for the first time on June 30. It accepts combinations of text, image and video as input and produces video output — priced at $0.10 per second of generated video, matching the rate for Veo 3.1 Fast. The current ceiling is 10 seconds of video per generation, with longer clips described as coming soon.

Google’s pitch is that the real value comes from chaining the two models together: use Nano Banana 2 Lite to generate an image quickly and cheaply, then pass that image to Gemini Omni Flash to animate it into a short video clip. The Interactions API handles session history across this pipeline, allowing up to three sequential edits in a single session. To make this concrete, Google published three remixable demo apps in AI Studio — Anywhere (transport yourself to landmarks in photos), Space Lift (reimagine interior designs), and Omni Product Studio (turn product images into e-commerce video clips) — that developers can fork and modify directly.

Know the Limitations Before You Commit

Gemini Omni Flash is still officially in public preview, and Google is transparent about what doesn’t work yet. Audio references are not supported. Video references up to 3 seconds are technically accepted by the API schema but are not correctly processed by the model. Character consistency across scene changes and panning movements is described as a known limitation. These are real gaps, not fine print — if your project depends on any of these capabilities, Omni Flash is not production-ready for that use case today.

How It Stacks Up Against Competitors

On the image side, Nano Banana 2 Lite’s $0.034-per-image price sits in the mid-range of a market that spans from roughly $0.009 to $0.24 per image. Google’s own Imagen 4 Fast and xAI’s Grok Imagine come in cheaper at $0.02 per image, and OpenAI’s GPT Image 1 Mini can go as low as $0.005 per image at lower resolutions. Where Nano Banana differentiates is its native integration with conversational, multimodal workflows — if you need to refine an image through a back-and-forth text conversation or pipe it directly into an agent pipeline, the Gemini ecosystem has structural advantages that pure image-generation APIs don’t replicate easily.

On the video side, Gemini Omni Flash enters a market recently reshuffled by OpenAI’s decision to discontinue the Sora web experience in April 2026, with the API set to follow in September. That’s a meaningful vacuum. At $0.10 per second, Omni Flash is priced at the budget end of the premium tier — below Sora 2’s $0.75 per second and Veo 3.1’s $0.15 per second fast-mode rate, and roughly on par with Kling 3.0. Its distinctive angle is conversational editing: the ability to refine video through natural language follow-up prompts, rather than regenerating from scratch each time.

Why This Matters If You’re a Student or Career Changer

Three things make this release worth paying attention to if you’re not a professional developer yet. First, Google AI Studio’s free tier — subject to rate limits that are explicitly described as generous enough for prototyping and classroom projects — means there is no financial barrier to starting. A full image-to-video pipeline is accessible today at zero cost.

Second, the remixable demo apps are actual working code, not tutorials. Forking an existing implementation, modifying prompts, and shipping a variation is a faster path to a portfolio piece than building from a blank file — and it’s exactly the kind of applied project that hiring managers in design, marketing and product roles are starting to recognize as a signal of practical AI competency.

Third, the skills this release rewards — prompt engineering, chaining multimodal APIs, managing cost-quality tradeoffs in generative pipelines — are not exclusively the domain of computer science students. Communications, UX and marketing students who learn to operate these tools fluently are building a genuinely differentiated skill set. When generating an image or a video clip becomes cheap and fast, the differentiator is judgment: knowing what’s worth making, for whom, and why. That’s not a purely technical question.

Both models ship with SynthID watermarking by default, and AI-generated content can be verified through the Gemini app, Chrome, or Search — a baseline transparency measure worth understanding if you’re building anything intended for public audiences.

Source: Google DeepMind