Models · 6 October 2026 · 1 min read
Nano Banana 2.1: Google’s image model gets better at text
Released on 6 October 2026: an image model that edits in conversation, writes clearer text in several languages and is already inside Google Ads.
By The Crewnema team
Google released Nano Banana 2.1 on 6 October 2026. It is an image model built on Gemini 3.6 Flash, so it reads text and pictures and answers with pictures and text. These are the points from Google DeepMind’s model card.
What Google published
- Image creation and editing with fine control and quick back-and-forth.
- Clear text for posters and diagrams, including text in several languages.
- Better consistency when several reference pictures are given.
- Availability in the Gemini app, Google AI Studio, the Gemini API, AI Mode in Search, Google Ads, Flow and Stitch.
The model card is unusually frank about limits: small text and long paragraphs can come out blurred, a character does not always match its input picture, edits by drawn mask are followed only partly, and it sometimes keeps the pose of the input or mixes up left and right.
Our view
Text in several languages is the feature to watch. A festival poster in Hindi, Tamil or Bengali is a common need for an Indian business and something image models have done badly. Its presence in Google Ads matters too: many businesses will meet this model there without choosing it. We would take the limits list at its word, though. Blurred small text means prices and phone numbers still belong on a card set as real text, not in the generated picture.
Side by side
In Crewnema
Nano Banana 2.1 is not on Crewnema’s roster. Your end card and any words you add on screen are set by Crewnema as real text, whatever model draws the picture: why that matters.
Sources
Make your first video
Crewnema is in private beta and lets businesses in a few at a time. Request access and we will write when yours is open.