Google Nano Banana 2.1 is out: better text and character consistency for product imagery
On 6 October 2026 Google made Nano Banana 2.1, the Gemini 3 family's image generation and conversational editing model, generally available. An update to Nano Banana 2, it raises quality while keeping the Flash tier's speed and cost profile.
Key improvements
- Quality and prompt adherence: Google says the model follows prompts more faithfully.
- Text rendering: Text inside images is more accurate, one of the biggest complaints in packaging and poster tests.
- Consistency: Characters stay more consistent across multi-turn edits. The model fuses up to 14 reference images, with consistency for up to 4 characters and fidelity for up to 10 objects.
- Resolution: 1K, 2K and 4K output, with tiling artifacts on wide and panoramic ratios fixed.
What changes for product visualisation?
Reference count and object fidelity matter most for brands. When a product's logo, label or pack pattern doesn't drift between scenes, AI-generated imagery can be used in campaigns with more confidence. Still, no model knows brand rules on its own. Anchoring on a real product photo and checking output against the brand guideline remain essential.
Our approach is unchanged: we anchor the product with a real shoot or 3D, build the scene with AI and do the final check with human eyes. Better models speed this flow up; they don't replace it.
If you want to talk about how your brand should adapt to these shifts, request a quote.
Source: Let's Data Science, October 2026
