Google Releases Nano Banana 2.1 for Image Generation and Editing
Google updated its image generation model in the Gemini API, addressing long-standing issues with text rendering and character consistency in panoramic images.
Deep dive
- Improves prompt adherence and text rendering accuracy in generated assets.
- Enables character consistency for iterative editing sessions.
- Resolves tiling artifacts specifically for 2K and 4K wide-format panoramic compositions.
- Model is accessible via the standard Gemini API endpoint.
Decoder
- Tiling artifacts: A visual defect in AI-generated images where the model repeats patterns or creates visible seams because it processed the wide image as smaller, disjointed squares.
Original article
Google released Nano Banana 2.1 in the Gemini API on October 6. The update focuses on image generation and conversational editing, with output at 1K, 2K and 4K resolutions.
What changes
Google reports better text rendering, prompt adherence and character consistency across successive edits. It also says the update fixes tiling artifacts in wide and panoramic images at 2K and 4K, including 1:4, 4:1, 1:8 and 8:1 aspect ratios.
For visual workflows, those changes target two practical problems: keeping a subject consistent through revisions and producing wide compositions without repeated image fragments. These are Google’s reported improvements; UX News has not independently tested them.
Availability
The model is generally available through the Gemini API as gemini-nano-banana-2.1. Google has deprecated the earlier gemini-3.1-flash-image model and recommends migration, but has not announced a shutdown date. The API release does not establish availability for every user of the Gemini consumer app.
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi