Nano Banana Pro not working? Missing, capped, refused or erroring: the fix for each

Nano Banana Pro not working? First check where you're using it, then check that it's really Pro. Nano Banana Pro is Google's Gemini 3 Pro Image model, and every Google app that runs it sets its own rules. In the Gemini app it isn't the model that draws your image: Google's help center makes it a redo option for paid plans, and once the day's Nano Banana 2 quota is gone, "you can’t redo any additional images with Nano Banana Pro." Google Flow caps Pro images per day and doesn't count failed or blocked ones. The Gemini API has no free tier for it, and the model's old preview name was retired in June 2026. Everything below comes from Google's and Higgsfield's own help pages, docs, status pages and terms, read November 7, 2026. We generated nothing for this page.

This page goes symptom by symptom: the Gemini app, Flow, Google AI Studio and the API, then Higgsfield. If your question is whether you can use Pro without paying, start with where Nano Banana Pro is free, and how many images you get; for the platform that offers it next to Kling and Seedance, see our Higgsfield guide.

Where are you using Nano Banana Pro?

Google runs Nano Banana Pro in three places: the Gemini app, Google Flow (its AI filmmaking studio) and the Gemini API with Google AI Studio. Platforms such as Higgsfield offer it next to other models. Each sets its own limits, so the first fix is knowing which one stopped you. This is what each one's own pages said on November 7, 2026:

Why can't you find Nano Banana Pro, or why did you get Nano Banana 2?

Because in the Gemini app, Pro isn't the model that draws your image. Google's image help article starts images on Nano Banana 2 and offers Pro to paid subscribers as a way to regenerate one, for extra detail, especially on text and infographics. The steps, per that article:

1. Be on a Google AI plan (Plus, Pro or Ultra). On Google's limits and features page, the Nano Banana Pro redo row is ticked only for those three plans.
2. Set the Gemini model to Pro before you ask for the image; Google's article ties the Pro redo to the Pro model.
3. Create the image. It comes from Nano Banana 2, or from Nano Banana 2 Lite when the model is set to Flash-Lite.
4. At the bottom right of the image, open More and choose Redo with Pro.

Two more things make Pro look broken in the app. Google's pages name the starting model differently: the help article says Nano Banana 2, while Google's US overview of Gemini image generation says a Google AI plan unlocks Nano Banana 2.1. Either way, Pro isn't the model that draws your first image, so check the model name under the image. And the model can change under you: Google's limits page says a subscriber who reaches the usage limit can keep going on Flash-Lite, and its image article says Flash-Lite requests use Nano Banana 2 Lite. By our reading, that means no Pro redo until the limit refreshes and you switch back to Pro.

In Google Flow

Flow's own pages disagree on where you start. Its image help says the prompt box defaults to Nano Banana Pro, while its models page says Nano Banana 2 Lite "is the default model available at no charge" and calls Pro the default for Ultra subscribers. Flow's Agent mode is a third way to miss Pro: with the Agent on, Flow sends your request to whichever image model it judges best. To be sure, turn the Agent off, click the model name in the prompt box, choose Image and pick Nano Banana Pro.

If you'd rather not pay Google, Flow's daily allowance is the free route to Pro. How Flow's free Nano Banana Pro allowance works has the full limits table and the watermark rules.

Hit a Nano Banana Pro limit?

In the Gemini app

Gemini's limits are measured in compute, not in a fixed number of images. Google's limits page says "Your limit refreshes every 5 hours until you reach your weekly limit." Image, video and music requests use more of that allowance than chat. Gemini warns you when you're close and, once you hit the limit, tells you when it refreshes; Settings, then Usage Limits, shows where you stand. Google publishes no image count for personal plans.

Your options, per Google: wait for the refresh, or move to a plan with more headroom. Plus gets 2x the standard limits, Pro 4x, and Ultra 5x or 20x Pro's, depending on the subscription. On any plan, the catch in the screenshot above still applies: once the day's Nano Banana 2 quota is gone, Pro redos stop too, and Google gives no reset time for that daily image quota on personal accounts. Google also warns that when demand spikes, accounts without a plan may be limited first.

In Google Flow

Flow counts images per model per day, without spending its credits. Nano Banana Pro gets up to 20 a day without a Google AI plan, 100 on Plus or Pro and 500 on Ultra. The count resets 24 hours after your first image with that model, and if an image fails or a safety filter blocks it, Google says "it does not count toward your daily limit".

Two other Flow messages point to rate limits, not the daily cap. "You're requesting generations too quickly" means Flow is limiting how many generations you can start, especially on models that cost no credits, and Google's Flow FAQ says that per-minute allowance shrinks after a large number of generations in a day. If a generation fails with a message about unusual activity, Google's fix is to retry after a couple of minutes with any VPN or proxy turned off. A card stuck on Pending may need a manual retry or deletion; check the notifications in the top-right corner first.

On a work or school account

Google publishes separate image limits for Workspace accounts, and on several editions they're tight: three Pro redos a day. Its work and school accounts page notes that "Image generation & editing is in high demand" and that these limits change often and reset daily.

Work accounts must belong to someone 18 or older, editions that add Gemini as an additional service need the admin to switch it on, and on school accounts some features are for users 18 and over only. If you see no Pro, Expanded or Ultra badge at the top of Gemini, you're on standard access, and only your Workspace admin can change the license.

Why does Nano Banana Pro refuse your prompt or photo?

Google doesn't publish the wording of Gemini's refusals or a list of the people, brands or words that trigger them. What it does publish:

• The Gemini app: Google's image help says Gemini may remove images when its systems detect a possible breach of Google's Terms of Service or its Prohibited Use Policy, and asks you not to violate other people's copyright or privacy rights.
• The Prohibited Use Policy: Google's policy for its generative AI bars content that violates privacy or intellectual property rights (including using someone's biometrics without the consent the law requires), impersonating a real person, living or dead, without disclosure in order to deceive, sexually explicit content, and anything exploiting children.
• Flow: its FAQ says Flow has safeguards for minors and for photos of people you upload, and that some prompts won't generate. Its country page adds that features involving identifiable minors and prominent people vary by region.
• The API: a blocked image comes back with a code such as image_safety, image_prohibited_content or image_recitation (copyright), and Google's advice is to change the input and retry. In Google AI Studio, a blocked answer shows as No Content; hover over it and click Safety to see why.

Our advice, from those rules: describe what you want instead of naming a real person, brand or character; use photos you took or have permission to use; keep children out of edits; and if a harmless prompt keeps failing, send Google feedback (Flow has a feedback form, Gemini has thumbs up and down) instead of rewording it to slip past the filter. The policy lists "Circumvention of abuse protections or safety filters" among the things it prohibits. If the image is for a client or a paid ad, who carries the legal risk in AI client work on Higgsfield is worth reading before you put a real person or brand in it.

Is your account allowed to use it?

If nothing generates at all, check the account before the prompt:

• Gemini app, personal account: 13+ (or your country's age) to generate images and 18+ to edit them; a Google AI plan in the EEA, Switzerland or the UK needs you to be 18 or older.
• Google Flow: you must be 18 or older with a verified age, in one of Flow's supported countries. Google says a VPN won't get you in, and Flow isn't offered everywhere Google AI Pro is sold, so check Flow's country list before you pay for a plan to use it.
• Google AI Studio and the API: "You must be 18 or older to use Google AI Studio." A 403 Access Restricted error in AI Studio means a Terms of Service problem, and Google's AI Studio troubleshooting page names being outside a supported region as a common one.
• Work or school account: the edition table above applies, and your admin decides.

Nano Banana Pro errors in Google AI Studio and the Gemini API

Most API failures with Nano Banana Pro come down to four things: billing, the model name, rate limits and request parameters.

No free tier, so billing comes first

Google's Gemini API pricing lists Gemini 3 Pro Image with "Not available" in its Free Tier column, for input and output alike. To call it, you need the paid tier: per Google's billing page, that means linking a billing account and prepaying at least $5 of credits, and on a Prepay plan the API serves requests only while your balance stays above zero. A Google Cloud Welcome or free-trial credit "can't be used towards the Gemini API or AI Studio". Paid output costs $0.134 per 1K or 2K image and $0.24 per 4K image.

Billing problems show up as three errors: failed_precondition (400) can mean billing is off, payment_required (402) means the Prepay balance is used up, and permission_denied (403) can mean the project needs verification, billing or a payment. Google's troubleshooting page is blunt about those: "Do not retry on client errors (like 400, 402, or 403)". Fix the cause first.

Model not found? The preview name is gone

Nano Banana Pro launched in the API as gemini-3-pro-image-preview, and code written before mid-2026 may still call that name. Google's deprecations page lists the preview, released Nov 20, 2025, as shut down on June 25, 2026, with the stable gemini-3-pro-image as its replacement (released May 28, 2026, per the same page). A retired name returns model_not_found (404), and Google's error reference gives the fix: "Verify the model name or fall back to a different model."

429: which limit did you hit?

A 429 always means a limit, and Google's error reference splits it three ways: rate_limit_exceeded (too many requests or tokens per minute or per second), quota_exceeded (your daily quota) and too_many_requests (too many requests in a short time). Google's troubleshooting and rate-limit pages call the same status RESOURCE_EXHAUSTED. Three details explain most surprises:

• Limits apply per project, not per API key, so a second key in the same project shares the same limits.
• "Requests per day (RPD) quotas reset at midnight Pacific time."
• Image models also have an images-per-minute limit, and paid projects can have a spend-based limit over a rolling 10 minutes ($10 on Tier 1, depending on billing history) that also returns a 429.

For the per-minute kind, Google's fix is exponential backoff with a little random jitter; its Python SDK already retries up to four times by default. For the daily quota, wait for the reset or request a higher limit; your current limits are on AI Studio's rate-limit page.

Parameters that fail on Pro

• Size needs a capital K: 1K, 2K or 4K. Google's image docs warn, "Lowercase parameters (e.g., 1k) will be rejected."
• Inputs: Pro takes text and images. Video input works only on Nano Banana 2.1, Gemini 3.1 Flash Image and Gemini 3.1 Flash Lite Image, and no Nano Banana model takes audio.
• References: up to 14 images, of which Pro handles 5 with high fidelity.
• Image count: the model won't always return the exact number of images you ask for.
• No image at all: no_image means the model couldn't make one; a blocked image returns one of the image_ codes below.

Is Nano Banana Pro down right now?

For the API and AI Studio, Google keeps a status page at aistudio.google.com/status. It showed all systems operational on November 7, 2026, and its history includes two Nano Banana Pro incidents and one outage that hit every API key:

For the Gemini app on a work or school account, Google's Workspace Status Dashboard has a Gemini row (no incidents on Nov 7). We found no public status page for the Gemini app on personal accounts or for Flow linked from their help centers, so if the app fails for everyone, waiting and retrying is the only fix (our advice).

On Higgsfield, status.higgsfield.ai has an Image Generation component, at 100% uptime over 90 days on November 7. Its changelog, though, logged a rise in failed Nano Banana Pro generations on August 13, 2026, inside that window: "The cause was a temporary provider-side issue." Credits for those failed generations were refunded automatically. For one model, the changelog is the better record (our reading).

Nano Banana Pro stuck, failed or flagged on Higgsfield?

Higgsfield is a separate creative platform that runs Google's Nano Banana Pro on its own credits: open Image, then Nano Banana Pro, per its help center. Signed out on November 7, 2026, its Generate button showed 2 credits for one image at 1K and again at 2K, and its pricing table lists Nano Banana Pro at "2 credit/image", or 4 at 4K. The 4K option carries a Premium tag in the quality menu; signed out, we couldn't select it. When a job there misbehaves, Higgsfield's rules apply, not Google's:

• Stuck: a Queued job can be canceled from its tile in Assets, with the credits returned; a Processing or Generating job has to finish or fail. Credit generations run in the priority queue, while Unlimited and free ones share a pool that can be slower.
• Failed: "You get the credits back automatically, usually within a few minutes." Check Manage Account, then Usage. If a refund is missing, Higgsfield asks for the generation ID, the model, any error message and your browser and OS. Higgsfield's queue statuses and refunds, explained has the details.
• Flagged: Google's filter runs on Higgsfield too. Its help center says model providers, Nano Banana among them, "also have their own independent safety systems on top of Higgsfield's", that Higgsfield can't see those providers' rules, and that it can't reverse a classifier's decision; credits for NSFW-flagged jobs come back for most models. Why Higgsfield flags a generation as NSFW covers false positives and how to report one.
• Blocked for copyright or IP: Higgsfield's own check catches named characters, teams, artists and brand logos, in the prompt or in a reference image. Describe instead of naming; Higgsfield applies these blocks platform-wide and won't lift them for one account. Every other message is decoded in Higgsfield error messages.

Higgsfield's own fixes for Nano Banana Pro output

For images that come back wrong rather than failing, Higgsfield's Nano Banana article (last modified Sep 9, 2026) suggests:

If the problem is one face drifting from image to image, see how Nano Banana compares with Higgsfield's own Soul model, and our same-face test across eight scenes for what drifted on Nano Banana 2 (not Pro) and Seedream.

Another way to run it when Google's limits block you

Higgsfield offers Google's Nano Banana Pro on its own credits, next to other image and video models. That changes some of Google's limits and not others:

On Higgsfield's plan cards (signed out, Nov 7, 2026), Starter's 270 credits a month are listed as 135 Nano Banana Pro images, Plus's 1,200 as 600 and Ultra's 3,000 as 1,500. Plus and Ultra also list 7 days of Unlimited Nano Banana Pro at 2K, which Starter's card crosses out. Higgsfield's Unlimited help says that window runs in the standard queue, works only on higgsfield.ai and doesn't restart at renewal, so read how Higgsfield Unlimited works before you count on it. Higgsfield's plans article says free access covers a small set of models and doesn't include credits (the new accounts we made on Oct 31, 2026 did get a few; see what a free Higgsfield account gets), so steady Nano Banana Pro use there means a paid plan. What a credit works out to per Nano Banana Pro image, next to Google's API price, is in what Nano Banana Pro costs when you pay.

What we couldn't verify

• How many Nano Banana 2 images a day personal Gemini plans allow. Google gives image counts only for work and school accounts.
• Whether Redo with Pro appears with the Gemini model on Flash or auto. Google's article ties it to the Pro model, and we didn't sign in to check.
• Which image model Flow starts you on. Its image help and its models page disagree.
• Whether teens can generate images in the Gemini app. Google's help article says 13+ (or the local minimum age) can generate and 18+ can edit, while its Gemini image-generation overview page carries a plain 18+ note.
• Whether Google AI Studio's playground runs Pro without a paid key. We didn't sign in; the API pricing table lists no free tier for the model.
• Google's rules for real people and public figures in Nano Banana Pro images. Google publishes general policies, not a list, and doesn't publish the wording of Gemini's refusals.
• Higgsfield's 4K price on the Generate button, and which plans unlock 4K. Signed out, the Premium-tagged option couldn't be selected; the pricing table says 4 credits.
• Exactly which Google filter settings apply on Higgsfield. Higgsfield says the provider's system runs on top of its own and that it can't see those rules.
• Outages nobody posted. We found no public status page for the Gemini app on personal accounts or for Flow.

More on Nano Banana Pro and Higgsfield

• Every Nano Banana Pro answer on one page
• Is Nano Banana Pro free? Flow's daily allowance and what Pro costs
• Nano Banana Pro vs Higgsfield Soul: which to use, and is either free
• Same AI face, eight scenes: what drifted on Nano Banana 2 and Seedream
• Higgsfield slow or stuck: the queue and refunds, explained
• Higgsfield error messages, decoded
• Seedance not working? The fix for each problem
• Kling AI not working? The fix for each problem