Nano Banana 2 Lite: images in 4 seconds, prices nearly zeroed and videos on NotebookLM

Google has announced Nano Banana 2 Lite, the new image generation model intended to replace the original version of the family, and has previewed Gemini Omni Flash for video generation. Among the applications that immediately adopt this news is NotebookLM, with a new format called Short Video Overviews.

Nano Banana 2 Lite, officially named Gemini 3.1 Flash-Lite Image, generates an image in about 4 seconds at a declared cost of $0.034 for 1,000 images in standard resolution. Google describes it as the fastest and most economical model ever introduced in the Nano Banana family, designed for high-volume workflows such as advertising tests or quick creative iterations.

Compared to the original Nano Banana (Gemini 2.5 Flash Image), which officially becomes the "legacy" model of the line, the newcomer brings improvements on three fronts: contextual knowledge for realistic scenes and geolocalized mockups, character consistency across multiple subsequent generations, and the rendering of readable text directly in images, useful for localized advertising variants in different languages.

Gemini Omni Flash enters public preview

Alongside the image model comes Gemini Omni Flash, already showcased at I/O and now available in public preview on Google AI Studio and Gemini API. The price is set at $0.10 per second of video output at 720p, the same pricing as Veo 3.1 Fast.

The model allows conversational edits on existing videos: replacing a character, changing the lighting of a scene, or adjusting the framing, while keeping the original audio and video tracks intact. It supports mixed multimodal inputs, combining text, images, and video clips to guide the final generation.

Google notes, however, that there are still limitations present in the preview phase: the maximum duration remains set at 10 seconds per generation, automatic scene extension is not yet available, and video references up to 3 seconds, although accepted by the API schema, are not currently processed correctly by the model.

The two models can be used in sequence via the Interactions API, which maintains session history and allows concatenation of up to three successive edits on the same content. Google has released some demo apps to accompany this, including one that transforms static product shots into e-commerce style videos.

On the consumer front, the rollout covers AI Mode in Search, the Gemini app, Google Photos, Stitch, Google Flow, and Google Ads, in addition to NotebookLM. All generated content remains watermarked with SynthID, verifiable through the Gemini app, Gemini in Chrome, or Search.

For NotebookLM, the integration of Nano Banana 2 Lite translates into Short Video Overviews: vertical clips of 60 seconds, designed for smartphones, that condense a document uploaded by the user into a narrated explanation accompanied by educational animations. The format recalls the Cinematic Video Overviews introduced in March 2026, but with a quicker and more immediate cut.

The feature is not yet active for everyone: Google speaks of an arrival "in the coming weeks," without specifying markets or specific plans involved in the initial rollout. It remains to be seen whether the Short Video Overviews will also appear in AI Mode, where Nano Banana 2 Lite is already operational for other purposes.