Nano Banana 2.1: What Google Improved and How to Use It

Nano Banana 2.1 is Google's latest image generation and editing model, and if you've used Nano Banana 2 before, the short version is: same job, cleaner results, especially on text, infographics, and keeping a character looking like themselves across edits. Nano Banana 2.1 is Google DeepMind's image generation and editing model, released on October 6, 2026 in the Gemini app, AI Mode in Search, Google AI Studio, the Gemini API, and Google Cloud. It's not a separate product you have to go hunting for — it's the model doing the work wherever Google already made images.

What Nano Banana 2.1 actually is

Google has released Nano Banana 2.1, a new model for image generation and editing. It replaces Nano Banana 2, released in February 2026, and Google says it improves on previous versions "across the board." It sits on Google's fast, lower-cost tier rather than the top-end Pro tier: Nano Banana Pro likely remains Google's best image model, but it uses the most recent Pro model, 3.1, while 2.1 uses Gemini 3.6 Flash.

That matters for how you should think about it. This isn't Google's "best possible image, no matter the cost" model — it's the everyday workhorse: good enough for most jobs, fast, and noticeably cheaper to run than its predecessor. Google on Tuesday released Nano Banana 2.1, the newest version of the AI model that turns a typed sentence into a picture, or edits a photo you already have. It's rolling out in the Gemini app, Google Search's AI Mode, Google Ads and developer tools including Google AI Studio, Flow and Stitch.

What's actually improved over Nano Banana 2

Google didn't ship this as a full redesign — it's a focused update, and the focus is visible in four places.

Text and infographics. This is the headline fix. According to Google, version 2.1 improves visual quality, text rendering, character consistency across conversation turns, wide panoramas, and infographics. If you've ever asked an image model for a chart, a labeled diagram, or a poster with real body copy and gotten garbled letters back, this is the category Google targeted directly. Independent coverage of the model card agrees: Google's model card shows clear gains in infographic design and accuracy, multi-character consistency, and mask and sketch editing.

Seamless wide panoramas. Nano Banana 2 had a known problem on very wide images — visible seams or repeating patterns when you stretched a scene across an ultra-wide frame. Nano Banana 2.1 adds a medium thinking level, now the default, and fixes the tiling Nano Banana 2 could show on very wide frames at 2K and 4K. If you make banner images, letterbox cinematic shots, or wraparound backgrounds, this is the fix that was missing before.

Multi-turn character consistency. This is the feature that matters most if you're building anything with a recurring character — a mascot, a product across a campaign, a consistent face for a story. It can process up to 14 reference images at once, keeping up to four characters and ten objects consistent. Google's own documentation is specific about the numbers: Google's documentation splits that into up to 4 characters, whose faces and outfits stay consistent, and up to 10 objects, which keep their shape and detail. That's a real ceiling, not marketing language — if you're trying to keep five characters straight across a scene, expect the fifth one to drift.

Mask-based editing and subject consistency. Google frames this as the other half of the release, alongside the raw generation gains: Google points to three upgrades: better visual design, sharper mask-based editing (you mark one region and only that area changes), and better subject consistency, meaning a person or object still looks like itself after several edits. In practice this means you can paint over just the sky, just the jacket, or just the background — and the rest of the image stays untouched, instead of the whole picture subtly re-rolling.

Thinking levels. New to 2.1 is a tunable reasoning step before the image renders. The model also connects to Google Search and offers minimal, medium, and high thinking levels that affect image quality. Higher thinking generally means better adherence to a complex prompt — more objects, more text, more spatial instructions — at the cost of a slightly longer wait.

What Nano Banana 2 users should know

Google will shut down the previous model, "gemini-3.1-flash-image," on October 29, 2026. Reporting on the exact cutoff has been inconsistent — some trackers note Google later withdrew that specific date — but the direction is clear either way: Nano Banana 2 is deprecated, and new work should move to 2.1.

Price, for context

Google cut the per-image API cost roughly in half with this release. Through Google's developer API, the pipe that lets other companies' apps use the model, a standard 1K image costs $0.0336. At the standard 1K price, a thousand images now cost about $33.60 through the API, down from $67 on Nano Banana 2. Resolution and reference handling also scaled up: Nano Banana 2.1 is built on Gemini 3.6 Flash, with 1K to 4K output, up to 14 reference images, panoramic aspect ratios and API pricing from $0.0336 per image.

Do it in Plurel

You don't need a Google developer account or an API key to use any of this. In Plurel, Nano Banana 2.1 is just a model choice inside the Image tool, on the same credit balance as everything else.

1. Open Plurel and sign in with Google — free, no card required.
2. Click Image in the row under the composer, or go straight to plurel.tech/?model=nano-banana-2.
3. Pick Nano Banana 2.1 from the model list. It costs 4 credits per image. (You can also open its page directly at plurel.tech/nano-banana-2-1.)
4. Type your prompt. For the features above, be specific:
   • For an infographic: describe the layout, the exact headline text, and how many data points or icons you want.
   • For a panorama: ask for a wide aspect ratio explicitly and describe the full span of the scene, left to right.
   • For a consistent character: upload 1–4 reference images of the same person or object and describe the new scene or outfit.
5. Generate. If you like the result but want one thing different, just type a follow-up in the same chat — "same character, now on a rooftop at night," or "make the sign say CLOSED instead" — and Plurel edits what was just made rather than starting over.

For targeted changes to a photo you already have — removing an object, swapping a background, fixing one region without touching the rest — use the Edit tool instead, at plurel.tech/?model=image-edit. It applies the same mask-based editing strength Google built into 2.1, so you mark what should change and the rest of the image stays put.

If a chat model is planning your visuals for you — Claude Opus 5, Claude Sonnet 5, or GPT-6 Astra — it can pick Nano Banana 2.1 as part of a plan and show you the credit cost before anything generates.

Tips and troubleshooting

If text still comes out wrong: keep the text short and put it in quotation marks in your prompt ("a sign reading 'OPEN 24 HOURS'"). Longer sentences of body text are still the hardest case for any image model.

If a character drifts after several edits: you're likely past the model's four-character ceiling, or asking for too large a change in one step. Make changes incrementally — one edit at a time — rather than rewriting the whole scene at once.

If you need photorealism over design accuracy: Nano Banana 2.1 is the fast, efficient tier, not Google's top-end model. For the highest realism Google offers, compare it against other options in our best AI image generator comparison, or try GPT Image 2.5 or Seedream 5 side by side on the same prompt.

If you want the image to move: Nano Banana 2.1 only makes stills. Generate the image first, then send it to Animate to turn it into a video clip.

For background on how keeping a face or product steady across images actually works under the hood, see Image to Video Consistency and Consistent Character Prompt.

The short version

Nano Banana 2.1 is Google's October 2026 refresh of its fast image tier — sharper text and infographics, seam-free wide panoramas, steadier characters across edits, and a cheaper price per image than before. In Plurel it's one click in the Image tool for 4 credits, with the Edit tool standing by for anything you want to change afterward.