Google launches fast image model and new video model with audio

Google has launched two new AI models that can create images and video from text. They are called Nano Banana 2 Lite and Gemini Omni Flash, and they were unveiled on June 30, 2026.

Images in a few seconds

Nano Banana 2 Lite, also known as Gemini 3.1 Flash-Lite Image, is Google's fastest and cheapest image model in the Nano Banana family, according to IT Brief. It can both create and edit images.

According to Google, the model can generate an image from a text description in about four seconds. Editing takes a little longer. The price is $0.034 per image at 1K resolution, according to Google's pricing.

Google highlights that the model is good at following instructions, keeping characters consistent from image to image and rendering text so that it is legible. It is aimed at rapid ideation and fast-paced work.

The model is available in Google AI Studio, the Gemini API and Gemini Enterprise Agent Platform. Google also plans to roll it out in its own products such as AI Mode in Search, the Gemini app, NotebookLM, Google Photos and Google Ads.

Video that can be edited through conversation

Gemini Omni Flash creates video and can edit it in a conversation where users type what should be changed. It takes text, images and video as input. According to IT Brief, it can swap characters and products, transfer styles, add objects and change lighting, among other things. It also generates audio along with the video.

The model costs $0.10 per second of generated video. It is in public preview, so it has not been formally released, and it is available in Google AI Studio, the Gemini API, Gemini Enterprise Agent Platform, the Gemini app and Google Flow.

There are several limitations. Videos can last no more than 10 seconds, though Google says this will be extended. The model processes video references of up to 3 seconds. In the API, audio references and scene continuation are not supported, and characters may change appearance during scene cuts or panning.

Labeling and partners

Content from both models gets an invisible SynthID watermark intended to show that it was made with AI. At Google Cloud, C2PA content credentials are also included by default.

Adobe is integrating both models into Firefly, and WPP has built Omni Flash into its marketing platform WPP Open. Figma is testing Nano Banana 2 Lite, and Invideo, Artlist and Manus AI are also testing the models.

WPP's CIO, Elav Horwitz, says Omni Flash's ability to work with multiple types of input is a leap forward for controlled AI production. Artlist's Idan Yonas says: "Speed is no longer a limitation. When generation is faster than imagination, creators stay inside the idea."

The claims about speed and quality come from Google and its partners, not from independent tests.