What to know
- Nano Banana Pro is Gemini 3 Pro’s new image generation and editing model, focused on studio-quality control, accurate text, and richer world knowledge.
- Access is available inside the Gemini app via the “Create image” option with the Thinking model, with a limited free tier before usage shifts to other plans.
- The model supports detailed editing, multi-image compositions, and embedded provenance metadata (C2PA and SynthID) to help identify AI-generated images.
- Google is rolling it out across Gemini, NotebookLM, Workspace, Google Cloud, and Ads, while keeping the original Nano Banana model for lighter, faster tasks.
Nano Banana began as Gemini 2.5 Flash Image, a fast, prompt-friendly generator optimized for one-shot edits and social-ready visuals. The Pro version, built on Gemini 3 Pro, shifts the focus toward precision, consistency, and deeper reasoning over more complex visual tasks.
The original model drew attention for hyperrealistic 3D figures and flexible image editing with simple text prompts, from restoring photos to creating mini figurines. Nano Banana Pro keeps these strengths but adds stronger understanding of context, layout, and subject matter, making it better suited to infographics, diagrams, and professional design workflows.
How and where to try Nano Banana Pro for free
Nano Banana Pro is live inside the Gemini app. Select the Thinking model.
And choose “Create image”.
This will routes image generation through this new engine. On the free tier, generation quotas are limited, after which requests may fall back to the original Nano Banana model unless a paid Google AI subscription is in place.
Beyond the consumer app, Pro is rolling out to NotebookLM, Google’s research and writing assistant, and into Google’s developer and enterprise stack via the Gemini API, Vertex AI, and related tools. In Search, Nano Banana Pro integrates with AI Mode for US users on Google AI Pro or Ultra, and it is set to appear in Flow, Google’s AI filmmaking tool, for Ultra subscribers first.
What the 'Pro' changes
Nano Banana Pro is built on Gemini 3 Pro Image, leveraging the language model’s reasoning and real-world knowledge to visualize structured information, like live weather or sports scores, inside a single image. This allows generation of contextually accurate infographics, data diagrams, and multi-panel visuals that stay aligned with current information, though hallucinations remain possible.
The Pro model significantly improves text handling, rendering legible, stylized text across many fonts and languages, which is key for posters, invitations, ads, and slide titles. It also adds greater control over composition and layout, so complex prompts about where to place elements or how to structure a diagram translate more reliably into the final image.
Image generation capabilities and styles
Nano Banana Pro supports high-resolution image generation up to 4K across multiple aspect ratios, helping assets translate cleanly to print, presentations, and large displays. Compared with the original Nano Banana’s 1024–2048 pixel outputs, this increase in resolution gives more room for fine details, especially in typography-heavy assets.
The model handles a range of styles, including realistic portraits, product renders, cinematic scenes, and stylized illustrations, while maintaining character and scene consistency across iterations. For creators used to multi-turn prompting, Pro is designed to preserve lighting, proportions, and color palettes as edits accumulate, which matters when building branded sets or story-driven sequences.
Multi-image, multi-person and composition control
Nano Banana Pro can blend up to 14 separate images or objects into a single composition, which helps when combining product shots, icons, screenshots, or reference sketches. It is tuned to maintain resemblance and consistency for up to five different people in an image, which is useful for group shots or narrative scenes with recurring characters.
The model’s layout understanding means prompts describing positions, camera framing, or relationships between objects (such as “a chart in the foreground, city skyline in the background”) translate with more reliability than before. This level of compositional control is particularly important for infographics and slide content, where hierarchy and spatial organization need to be clear at a glance.
Advanced editing features and fine-grained control
A central focus for Nano Banana Pro is editing existing images, not just generating new ones. The model supports targeted, local edits such as changing a subject’s pose, removing objects, or adjusting specific regions while leaving the rest of the image intact.
Users can adjust camera perspective, simulate bokeh, refocus areas, tweak color grading, or flip lighting from day to night for cinematic effects. Combined with high-resolution output, this makes the model suitable for retouching marketing content, refreshing product photography, or testing multiple visual directions from the same base shot.
Safety, watermarking and provenance
To address concerns about deepfakes and synthetic media, Google embeds both C2PA metadata and its SynthID invisible watermark into images created or edited with Nano Banana Pro. C2PA tags are designed to be interoperable across platforms, which should help future tools and social networks detect and label AI-generated images more reliably.
SynthID allows Gemini to analyze an uploaded image and indicate whether it was generated by Google AI, even if it has been resized or lightly edited. TikTok and other platforms are beginning to support similar metadata, suggesting that Nano Banana Pro’s watermarking will tie into a broader ecosystem for provenance and authenticity.
Discussion