Google Releases Nano Banana 2.1 with Improved Mask Editing and Subject Consistency
Google has officially released Nano Banana 2.1, an incremental update to its generative AI image model. The new model improves overall visual composition, precise mask-based regional editing, and subject consistency across multiple generated images. Subject inconsistency and full-image regeneration have long been major pain points for creators and businesses using generative AI image tools. By integrating Nano Banana 2.1 across its ecosystem—including Gemini, Google AI Studio, and Google Ads—Google enhances practical usability for professional workflows. Developers can access the model via the ID `gemini-nano-banana-2.1`, which accepts multimodal inputs (text, images, audio, and video) and outputs images in 1K, 2K, and 4K resolutions. Mask editing enables users to modify isolated sections of an existing image without altering unselected regions or triggering quality degradation.
## BACKGROUND
Subject consistency in generative AI refers to maintaining the visual identity of a character, mascot, or object across different scenes, poses, and prompt iterations. Mask editing, or inpainting, lets users define specific areas of an image for the AI to redesign while leaving the surrounding context intact. Earlier in the year, Google introduced Nano Banana 2 (Gemini 3.1 Flash Image) to bring high-quality output into its high-speed Flash model tier.