Summary: On June 30, Google released Nano Banana 2 Lite as gemini-3.1-flash-lite-image, positioning it as the fastest and cheapest Nano Banana model at about four seconds and $0.034 per 1K image. Gemini Omni Flash also entered public preview at $0.10 per output second, matching Veo 3.1 Fast. Both outputs carry SynthID watermarking, and Lite remains available in AI Studio, the Gemini API, and the Agent Platform.
Google used its blog on June 30 to roll out Nano Banana 2 Lite, published under the model name gemini-3.1-flash-lite-image. The company frames it as the fastest and most cost-efficient image model in the Nano Banana family: text-to-image results arrive in roughly four seconds, a 1K-resolution image runs $0.034, and the package targets high-throughput, rapid-iteration workloads. Alisa Fortin and Anish Nangia, product managers at Google DeepMind, are credited on the post, which also brought word that the Gemini Omni Flash video generation model is entering public preview at $0.10 per second of output — priced identically to Veo 3.1 Fast.
The notable part is not any single model but the standardized handoff between them: produce stills quickly with Lite, then bring those stills to life with Omni. Google puts it bluntly — “the real magic happens when you chain these models together.”
The Math Behind 3.4-Cent Images
A price tag of $0.034 rewires habits. With a draft image running about a cent, spinning up ten candidates and keeping the best stops being an indulgence and becomes the standard loop of interactive prototyping. The arithmetic scales handsomely: a thousand drafts cost $34, folding aggressive exploration into budgets almost any content or product team already carries. Google is careful to add that Lite retains character consistency, prompt adherence, and legible in-image text even with the speed emphasis — the trio quick-draft pipelines tend to give up first.
The original Nano Banana (Gemini 2.5 Flash Image), meanwhile, has been tagged as legacy, with Google explicitly recommending a move to Lite. Two years ago it was the flagship image capability; now it fills the entry-level slot. Few sharper footnotes exist for how far image-generation costs have dropped.
Lite, 2, and Pro: Three Tiers, Three Jobs
Image generation at Google now runs on three tiers. Speed and scale sit with Nano Banana 2 Lite; everyday generation belongs to Nano Banana 2, the balanced “workhorse” of the lineup; and Nano Banana Pro guards the quality ceiling for demanding professional work. Nano Banana 2 arrived in February billed as “Pro-level capability, Flash-level speed,” and Lite takes the cost frontier one more step down, leaving price bands and use cases across the family cleanly divided.
The developer-facing consequence is easy to spell out: send drafts, bulk generation, and final output to different models, and manage three cost structures through one API.
Gemini Omni Flash: Chaining Stills into Motion
Omni Flash (gemini-omni-flash-preview) combines video generation with conversational editing and debuted at Google I/O. It takes text, image, or video as reference inputs, draws on Gemini’s real-world knowledge, and can align text and graphics with actions inside the video. With conversational editing, you adjust the result in natural language rather than rebuilding the prompt from zero — which is why the path from still to moving image feels like one unbroken workflow. The constraints deserve equal weight: a single generation tops out at ten seconds, audio reference uploads and scene extension are not supported yet, video references are still handled incorrectly, and character consistency can slip across scene changes and camera pans.
That chain takes physical form in three demo apps built for remixing. Omni Product Studio converts product stills generated with Lite into cinematic e-commerce videos; Anywhere and Space Lift complete the trio. An Interactions API allows up to three edits in sequence per session, so image, edit, and animate can all happen within one flow.
Availability, SynthID, and the Slop Argument
Lite can be used today in Google AI Studio, the Gemini API, and the Gemini Enterprise Agent Platform, and it will spread from there to consumer touchpoints such as AI Mode in Search, the Gemini app, NotebookLM, Google Photos, Stitch, Flow, and Google Ads. Omni Flash is in public preview across AI Studio, the API, and the Agent Platform, with the Gemini app and Flow also included. Anyone who has watched earlier Gemini generations launch will recognize the rhythm: API access reaches developers and enterprise users first, and consumer products take the model in over the following weeks.
SynthID watermarking is baked into both models, and verification works through the Gemini app, Chrome, and Search. The moment matters for context: numbers cited by TechCrunch put about 60 percent of TikTok videos in AI-generated territory already, and the $75 million partnership between Google and A24 met fan pushback. Once generating an image costs an order of magnitude less, whether this “ten times cheaper” capability hardens into a slop factory is the question that will shadow this product line for the long haul.