
Realism at every size
Visual quality and realism improve across 1K, 2K and 4K output. Skin, thread and fabric hold up when you zoom into the 4K file.
Up to 4K · sharper lettering · frames from 8:1 to 1:8
Nano Banana 2.1 is Google's image model from 6 October 2026: sharper text, clean wide frames and up to 4K. The generator below runs it, and signed out you start free on GPT Image 2.5 with no account.
Nano Banana 2.1 came out on 6 October, and it is already in this generator at 30 credits a picture.
Both pictures came out of the generator on this page at 2K. The first is a prompt @churvikv shared on X, run word for word. The second is one edit of it: winter came, and the swans, the moon, the cottage and the brushwork stayed where they were.


Drag to compare before and after
It is Google's update to Nano Banana 2 (Gemini 3.1 Flash Image): the same fast Flash tier, with better pictures, better lettering and steadier faces. In the Gemini API it is called gemini-nano-banana-2.1.
The generator here runs the model itself, text to image and photo edits, at 1K, 2K or 4K. Signed out, start free on GPT Image 2.5 or the Free engine tab; Nano Banana 2 and Nano Banana Pro stay one pick away.
Every capability on this page traces to Google's Gemini API model page for 2.1 or its launch coverage, and every price to a published rate card.
Creative engine
Google positions it as its high-efficiency model for image generation and conversational editing. It keeps Nano Banana 2's speed and cost class and raises quality, prompt adherence, text rendering and character consistency.
Google rolled the model out across the Gemini app, AI Mode in Search, Google AI Studio, Flow, Stitch, Google Ads and the Gemini Enterprise platform, and to developers through the Gemini API.
Nano Banana 2.1 runs from an account at 30 credits a picture. Without one, this ChatGPT image generator still gives you a picture: one 1K run of GPT Image 2.5, the model behind ChatGPT Images, and drafts on the Free engine tab.

Signed out, pick GPT Image 2.5 in the model list and your first 1K picture comes back before anyone asks for an email. It is the same run the ChatGPT image generator on our homepage opens with.
Every 2.1 run is 30 credits at any size, edits included, shown before you press Generate. Nano Banana 2 is 50 and Pro is 70 on the same balance, and a failed run is refunded.
When you run 2.1 here it is Google's model itself, at the size you pick. For free drafts with no account, the Free engine tab runs our own model, Veida AI Image v1.
Pro ($89.99 a month) and Max ($149.99) run it with no credit cost per image and no cap on how many: every retake, every variation, every size.
Five improvements Google lists over Nano Banana 2, in Google's terms.

Visual quality and realism improve across 1K, 2K and 4K output. Skin, thread and fabric hold up when you zoom into the 4K file.
Tiling artifacts on wide and panoramic aspect ratios are fixed, and the ratio list runs out to 4:1, 8:1, 1:4 and 1:8 for banners, headers and tall phone stories.
Lettering, labels and infographic layout render more accurately, so the words in your poster read the way you typed them.
Multi-turn character consistency for up to four characters, and object fidelity for up to ten objects, across a sequence of edits.
Google's per-image API rate is $0.0336 for a 1K picture, $0.0504 for 2K and $0.0756 for 4K, roughly half of Nano Banana 2. Here, a 2.1 run is 30 credits against 50 for Nano Banana 2.
Posted on X by people who named the model. The eye is also a template in the generator above: tap it and the prompt is already written for you.

One sentence of prompt, and the iris comes back with its fibres, veins and a window caught in the pupil. Made on Magnific, prompt included in the template.

A sunlit table, a head tilted mid-laugh and dust glowing in the light, posted as a gemini-nano-banana-2.1 output.

Bold colour blocks, crisp edges and nothing smudged: the look for a sticker, a poster or an app screen.

An astronaut making a call from an old phone booth on a red planet, the ship hanging overhead. Made on getimg.
The comparisons people are searching this week, answered by what each model is for and what it costs here.
2.1 is the newer Flash model and the one Google improved: realism, text, wide frames and faces across edits. Here it costs 30 credits a run against 50 for Nano Banana 2.
In Google's own tests, 2.1 in thinking mode outperforms Pro on several measures, including overall preference, infographic design and multi-character consistency. Nano Banana Pro is 70 credits a run here; 2.1 is 30.
Pick 2.1 for 8:1 banners, poster lettering and photo edits that keep the rest of the frame, at 30 credits a run. GPT Image 2.5, OpenAI's newest, is one pick away on the same account at 50.
Nothing to install, and no step where you are asked who you are.
Tap the eye, the swan lake or the harbour portrait to load its prompt, or describe your own subject, light, framing and finish.
A 2.1 picture is 30 credits at 1K, 2K or 4K, in any ratio from 8:1 to 1:8. Signed out, start with one free GPT Image 2.5 picture instead.
When a picture is close, upload it back with up to ten reference photos and say what to change. The swan lake pair above is one such edit.
Google released it on 6 October 2026 as a stable model, rolling it out across Google's products and the Gemini API that week. Some Gemini and Flow accounts showed it a day or two before others.
Here it is 30 credits a run at any size, with no per-image cost on Pro and Max. Signed out, the free picture on this ChatGPT image generator is one 1K run of GPT Image 2.5. Inside Google's apps, 2.1 follows your Google plan's limits and needs a Google account.
In the Gemini app, Google AI Studio, Flow, Stitch and AI Mode in Search on a Google account; through the Gemini API as gemini-nano-banana-2.1; and on this page from any account, at 30 credits a run.
Google's API rate is $0.0336 for a 1K image, $0.0504 for 2K and $0.0756 for 4K. Here it is 30 credits a run at any size, shown before you press Generate, and a failed run is refunded. On Pro and Max there is no per-image cost at all.
On Google's own published tests, 2.1 with thinking outperforms Nano Banana Pro on several measures, including overall preference, infographic design and multi-character consistency. Here it also costs 30 credits a run against Pro's 70.
Yes. Upload a photo, up to ten per run here, and describe the change. The model keeps what you did not mention, as the swan lake pair above shows.
1K, 2K and 4K, at 1K by default. Ratios run from square and the usual 16:9, 9:16, 4:3 and 3:2 out to 21:9, 4:1 and 8:1 wide and 1:4 and 1:8 tall.
Free pictures, the signed-out GPT Image 2.5 run included, carry a small veida.ai mark in one corner. A Nano Banana 2.1 run on credits or a plan downloads as the clean original.
Credits are shared across everything Veida AI runs. You see the cost before each run, and a failed job is never charged.
Free · 1 free image now, no account
Plus unlimited runs on the free engine — no card, nothing to cancel.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
Charges appear as “VEIDA AI” on your card statement. You can cancel any time from Settings → Billing; cancellation takes effect at the end of the period you already paid for. Operator details, the full model list, and the refund window are on the about page.
Pick a template, or bring your own prompt and your own photo: 30 credits a picture, or no per-image cost on Pro and Max.