Nano Banana 2.1 for a Restaurant Menu Board: Draft the Layout, Check Every Price

Use Nano Banana 2.1 for the layout. Don't use it as the place your prices live. Google made Gemini Nano Banana 2.1 (model code gemini-nano-banana-2.1) generally available in the Gemini API on October 6, 2026, and says it renders text better and handles wide 4:1 images better than the model it replaces. That makes a long, narrow menu board worth trying. It doesn't make the prices on the image trustworthy. Lock the menu text in a source sheet first, ask for a layout that uses only that text, then check every price, name and word on the draft against the sheet. If the image keeps getting text wrong, keep the background design and set the prices yourself in an ordinary design program where the text stays editable.

This is a workflow built from Google's documentation, read October 7, 2026. No image was generated for this article and no Gemini account or model was tested. The bad drafts described below are written test cases to check against, not mistakes anyone observed.

Three things to settle before you plan around it:

• Access. Google's terms require you to be 18 or older, in an available region, and using the service for professional or business purposes. Google's model page links to the model in Google AI Studio; whether your account can use it there isn't verified here. Open the link and see.
• Cost. Google's pricing page lists no free tier for this model in the Gemini API. At the standard paid rate, image output alone works out to $0.0336 per 1K image or $0.0504 per 2K image; input and text/thinking tokens are billed on top. Those are API rates, not a quote for what AI Studio charges your account.
• Accuracy. Better text rendering is Google's description, not a promise that every digit comes out the way you typed it. A person checks every line before anything is printed or put on a screen.

What Google released on October 6

One more line from the same guide is worth reading before you choose a model: in its tips for accurate text in images, Google points to Gemini 3 Pro Image for professional asset production. Google calls Nano Banana 2.1 the more efficient counterpart to that model. For a board you'll draft, check and finish yourself, that may be fine. For finished artwork with no editing step, Google's own documentation steers you elsewhere.

Before you open anything: who can use it, what it costs, what Google keeps

Who can use it

The Gemini API Additional Terms (last updated April 28, 2026) say you must be 18 or older, may only use the services in an available region, and that Google AI Studio and the Gemini API are for developers building for professional or business purposes, not consumer use. Google's available-regions page lists the United States, Canada, Mexico, the United Kingdom, Spain and Italy among many others; check your own country there. If AI Studio turns you away, that page gives three likely reasons: your region, being under 18, or not having verified your age on your Google account.

These docs cover the Gemini API and Google AI Studio. Don't assume the consumer Gemini app has the same model, sizes or controls.

What a draft costs at the standard API rate

Prices below are from the Gemini API pricing page (last updated October 7, 2026), standard paid tier, in US dollars. Token counts for the 4:1 sizes come from the image generation guide.

On top of that: input is $1.50 per 1M tokens for text, image or video, and text and thinking output is $7.50 per 1M tokens. Those depend on your prompt and on how much the model thinks, so they aren't estimated here. Google Search grounding is billed separately once you pass 5,000 free search requests a month (shared across Gemini 3.x models), which is one more reason to leave it off.

Worked example: three layout drafts plus two corrections, all at 2K, is 5 × $0.0504 = $0.252 in image output charges at the October 7 standard rate. That is not the total cost of a board, and it isn't what AI Studio charges. It's the image-output line only.

4K is left out on purpose. The image guide lists 2,520 output tokens for a 4K image; the pricing page says a 4K image uses 3,780. Until Google's pages agree, no 4K budget is quoted here.

What Google keeps

The Additional Terms split this by how you use the service:

• Unpaid use, such as AI Studio without billing or free API quota: Google uses what you submit and what it generates to improve its products, and human reviewers may read it. The terms say not to submit sensitive, confidential or personal information.
• Paid use: the API counts as paid when you use it through a Cloud project with an active billing account. AI Studio counts as paid when your account has access to a Cloud project with active billing, or is a Workspace enterprise account. Google doesn't use paid prompts or responses to improve its products, but it still logs them for a limited period to detect policy violations and for required legal disclosures. Paid isn't the same as nothing kept.
• EEA, Switzerland and the UK: the paid-service data terms apply to all use there, including free AI Studio use.
• Search grounding, if turned on, has Google store prompts and output for 30 days.

Whatever your tier, put nothing in the prompt that you wouldn't print on the wall: no food costs, supplier names, staff or guest information, or menu changes you haven't announced. The terms also say Google won't claim ownership of what you generate but may generate the same or similar content for others, so don't count on the design being unique to you.

Step 1: Lock the source sheet

The sheet is the only place menu facts come from. The model reads it; it never adds to it. Write it before you open any tool, and give it a version number so everyone knows which prices the board is supposed to show.

Here's a fictional sample. Corner Test Kitchen isn't a real restaurant, and these prices are made up for teaching. The conditions at the end summarize the access, data-use and cost points above, so keep them with any copy, including your own version of the sheet.

MENU BOARD SOURCE SHEET (fictional example, not a real restaurant)
Conditions at the end of this sheet go with every copy of it.
Restaurant:        Corner Test Kitchen
Board:             Lunch board above the counter, 48 in wide x 12 in tall
Applies to:        Lunch, dine-in and takeout, Main Street location only
Version / date:    v3 / 2026-10-07
Approved by:       ______________________  Date: ____________

BOARD TEXT (exact, in this order; nothing else goes on the board)
CORNER TEST KITCHEN
LUNCH
Jalapeño Melt  $12.50
Tomato Soup, 12 oz cup  $6.75
House Lemonade, 16 oz  $4.00

REQUIRED NOTICES FOR THIS LOCATION (owner supplies and approves)
[None written for this fictional sample. A real board may need some.]

LEFT OFF THIS SAMPLE ON PURPOSE
Allergen, calorie, nutrition or dietary claims, certifications,
discounts, tax statements, opening hours, QR codes, phone numbers,
website addresses.

CONDITIONS - KEEP WITH EVERY COPY OF THIS SHEET
Status:  Untested. Written from Google's Gemini API docs, read 2026-10-07;
         no image generated, no account or model tried. Every draft is
         unchecked until it passes the full check.
Model:   gemini-nano-banana-2.1 (Gemini Nano Banana 2.1), Gemini API or
         Google AI Studio. Not confirmed for your account or the Gemini app.
Access:  Google's terms: 18 or older, available region, professional or
         business use. Check your account and age verification first.
Inputs:  Nothing sensitive, confidential or personal, and no private
         business information (food costs, suppliers, staff or guest
         details, unannounced changes), on any tier. Your own sheet too.
Data:    Unpaid (e.g. AI Studio without billing): Google uses inputs and
         outputs to improve products; human reviewers may read them.
         Paid (API via a Cloud project with active billing; AI Studio with
         access to such a project, or a Workspace enterprise account): not
         used to improve products, but logged for a limited period - not
         zero retention. EEA, Switzerland, UK: paid data terms apply to all
         use. Search grounding on: prompts and output kept 30 days.
Cost:    Standard Gemini API, pricing page of 2026-10-07: no free tier.
         Image output only $0.0336 (1K) or $0.0504 (2K) per image; input,
         text and thinking tokens extra. Not an AI Studio quote. No 4K
         budget: Google's guide says 2,520 tokens, its pricing page 3,780.
Rules:   Grounding off. Upload only images you have rights to. Keep
         SynthID. Owner approves every word, price and notice. Recheck the
         whole draft after any change. Proof at real size and distance. If
         text keeps failing, set it as editable text in a design program.

That last block matters. This sample is not a complete menu for a real location. Depending on where you operate and what kind of business you run, your board may need notices this sample leaves out. Deciding which ones apply is your job, together with whoever advises you on local rules. It isn't something to ask an image model. If a notice is required, write it into the sheet word for word and check it on the draft like a price.

The guide's own limitations section supports doing it this way: Google says Gemini works best with text in an image if the text is produced first and then the image is requested with it. Here, the text comes from you.

Step 2: Pick the size before you write the prompt

Start from the physical board or the screen, not from the model.

A printed board

The fictional board is 48 inches wide by 12 inches tall. 48 ÷ 12 = 4, so the ratio is 4:1. Google's guide lists these 4:1 sizes for Nano Banana 2.1:

The height gives the same answer: 1024 ÷ 12 ≈ 85.3 at 2K. Whether 85 pixels per inch is enough depends on the printer, the material and how far away people stand, so ask the printer for their minimum at 48 × 12 inches. Don't pick a number yourself. The right ratio doesn't guarantee the board will be sharp or readable.

If 2K falls short, enlarging the image won't fix a wrong price or a soft digit. That's another reason to set the final prices in a design program: text typed there as live text isn't limited to the image's pixel count.

A screen

If the board is a screen, match the screen's shape instead. A 16:9 screen needs a 16:9 image; stretching a 4:1 design to fill it distorts every letter.

Google lists 1376 × 768 at 1K and 2752 × 1536 at 2K for 16:9. Those are close to 16:9 but not exact: 2752 ÷ 1536 ≈ 1.792, while 16 ÷ 9 ≈ 1.778. On an example 1920 × 1080 screen:

• Fit to height: the 2K image becomes about 1,935 pixels wide, so roughly 15 pixels get cropped across the two sides.
• Fit to width: it leaves about 8 pixels of empty bars.
• 1K: the image is smaller than that screen and gets enlarged.

Get the real resolution from the screen vendor, and keep all text well away from the edges.

Margins and crops

Printers and screens both trim or cover the edges. Ask the printer for their bleed and safe area, and ask the screen vendor whether the display crops. Then check that nothing on the draft, especially a final digit, sits near an edge.

In the API, the ratio and size are set with aspect_ratio and image_size; without them, Google says the model matches an uploaded image's size or makes a 1:1 square. Whether AI Studio shows matching controls on your account isn't verified here. After any draft, check the file's pixel size matches what you asked for (4096 × 1024 for 4:1 at 2K) and write it on the handoff sheet.

Step 3: The first prompt

Copy the whole block. The notes at the top are for you, not the model: paste only the part under the line, with the text from your own approved sheet swapped in. The prompt follows the structure in Google's text-in-images tip: the exact text, a described font style and the overall design.

NOTES FOR YOU - READ FIRST, DO NOT PASTE INTO THE MODEL
Status:  Untested. Written from Google's Gemini API docs, read 2026-10-07;
         no image generated, no account or model tried. Every draft is
         unchecked until it passes the full check.
Model:   gemini-nano-banana-2.1 (Gemini Nano Banana 2.1), Gemini API or
         Google AI Studio. Not confirmed for your account or the Gemini app.
Access:  Google's terms: 18 or older, available region, professional or
         business use. Check your account and age verification first.
Inputs:  Nothing sensitive, confidential or personal, and no private
         business information (food costs, suppliers, staff or guest
         details, unannounced changes), on any tier. Your own sheet too.
Data:    Unpaid (e.g. AI Studio without billing): Google uses inputs and
         outputs to improve products; human reviewers may read them.
         Paid (API via a Cloud project with active billing; AI Studio with
         access to such a project, or a Workspace enterprise account): not
         used to improve products, but logged for a limited period - not
         zero retention. EEA, Switzerland, UK: paid data terms apply to all
         use. Search grounding on: prompts and output kept 30 days.
Cost:    Standard Gemini API, pricing page of 2026-10-07: no free tier.
         Image output only $0.0336 (1K) or $0.0504 (2K) per image; input,
         text and thinking tokens extra. Not an AI Studio quote. No 4K
         budget: Google's guide says 2,520 tokens, its pricing page 3,780.
Rules:   Grounding off. Upload only images you have rights to. Keep
         SynthID. Owner approves every word, price and notice. Recheck the
         whole draft after any change. Proof at real size and distance. If
         text keeps failing, set it as editable text in a design program.
Paste only owner-approved text from your source sheet.

===== PROMPT TO PASTE: EVERYTHING BELOW THIS LINE =====
Design a wide restaurant menu board, aspect ratio 4:1, for a printed panel
48 inches wide by 12 inches tall. It hangs above a counter and is read from
across the room.

Use only the text between the two lines below, exactly as written: same
spelling, accents, capital letters, punctuation, prices and serving sizes.
Do not add, remove, translate, shorten or reword anything.
-----
CORNER TEST KITCHEN
LUNCH
Jalapeño Melt  $12.50
Tomato Soup, 12 oz cup  $6.75
House Lemonade, 16 oz  $4.00
-----
Layout: restaurant name at top left, LUNCH under it, the three items in one
row across the board, each price directly to the right of its item name and
as large as the item name.
Style: flat graphic design with simple shapes and two colors, dark green and
cream. Bold, plain sans-serif type with high contrast.
Keep a wide empty margin on every edge with no text near it.
The board contains only the text above, flat color shapes and empty space:
no food photos or drawings of food, no logos, icons, badges or symbols, no
QR codes, no web addresses, no phone numbers, no extra words.

Notes on that prompt:

• Grounding stays off. Google's examples turn on Search grounding by adding the google_search tool to the API request, so leave that tool out. If you see a search or grounding switch in AI Studio, leave it off.
• No food photos. A generated plate of food isn't your food. If you want real dishes on the board, use photos you own and place them in the design program.
• Logos only if you own them. If you upload your logo, the guide's reminder applies: upload only images you have the rights to. Then check it wasn't redrawn.
• No QR codes from the model. A QR code drawn by an image model isn't a reliable ordering link. If you need one, make it from your real URL in a QR tool, place it in the design program and scan it before printing.

Step 4: Check the draft against the sheet

Check the draft at full size, one line of the sheet at a time. Read every character; don't skim for the general shape. Fill in a row for every piece of text on the sheet, and add a row for anything on the draft that isn't on the sheet.

Mark Unknown when you can't tell, for example a digit that could be a 5 or a 6. Treat Unknown as Fix.

What a reject looks like

These are written test cases for the fictional sheet above, so you know what to look for. They are not errors observed from Nano Banana 2.1; no draft was made.

For readability, a monitor tells you very little. Print a section of the board at actual size, or put the file on the actual screen, and read it from where customers stand. Ask someone who hasn't seen the sheet to read the prices back to you.

Step 5: Fix one thing at a time, then check everything again

Google recommends iterating on images in a multi-turn conversation, and its editing template asks the model to change only one element and keep everything else the same. For a menu board, request one fix per prompt. As before, the notes are for you; paste only the part under the line:

NOTES FOR YOU - READ FIRST, DO NOT PASTE INTO THE MODEL
Status:  Untested. Written from Google's Gemini API docs, read 2026-10-07;
         no image generated, no account or model tried. Every draft is
         unchecked until it passes the full check.
Model:   gemini-nano-banana-2.1 (Gemini Nano Banana 2.1), Gemini API or
         Google AI Studio. Not confirmed for your account or the Gemini app.
Access:  Google's terms: 18 or older, available region, professional or
         business use. Check your account and age verification first.
Inputs:  Nothing sensitive, confidential or personal, and no private
         business information (food costs, suppliers, staff or guest
         details, unannounced changes), on any tier. Your own sheet too.
Data:    Unpaid (e.g. AI Studio without billing): Google uses inputs and
         outputs to improve products; human reviewers may read them.
         Paid (API via a Cloud project with active billing; AI Studio with
         access to such a project, or a Workspace enterprise account): not
         used to improve products, but logged for a limited period - not
         zero retention. EEA, Switzerland, UK: paid data terms apply to all
         use. Search grounding on: prompts and output kept 30 days.
Cost:    Standard Gemini API, pricing page of 2026-10-07: no free tier.
         Image output only $0.0336 (1K) or $0.0504 (2K) per image; input,
         text and thinking tokens extra. Not an AI Studio quote. No 4K
         budget: Google's guide says 2,520 tokens, its pricing page 3,780.
Rules:   Grounding off. Upload only images you have rights to. Keep
         SynthID. Owner approves every word, price and notice. Recheck the
         whole draft after any change. Proof at real size and distance. If
         text keeps failing, set it as editable text in a design program.
Paste only owner-approved text from your source sheet.

===== PROMPT TO PASTE: EVERYTHING BELOW THIS LINE =====
Using the provided image, change only the price next to "Jalapeño Melt"
from "$12.00" to "$12.50". Keep everything else in the image exactly the
same, including all other text, prices, colors and layout.

Then run the whole check again, every line, not just the one you fixed. A regenerated image can change text you didn't ask about, and nothing in Google's documentation promises an edit leaves the other lines alone.

Step 6: When to stop and finish in a design program

Stop correcting when the same line is still wrong after two fix requests, or when a fix breaks a line that used to pass. At that point:

1. Keep the layout you like. Ask for the same design with no text at all, leaving the panels and spaces where the text goes, and check that version for stray letters too.
2. Open it in an ordinary design program you or your designer already use, and type the approved text from the sheet as live, editable text.
3. Check the finished file against the sheet with the same table.

This is often the better finish anyway. When the soup goes from $6.75 to $7.25 next month, an editable file is a quick edit and a recheck of one line. A flat image means generating and checking the whole board again.

Keep the menu readable online, too. An image of the board shouldn't replace a text menu on your website; customers on phones need prices they can read and search. The first restaurant website checklist covers that, and if the board is crowded before design even starts, your opening menu may be too complicated.

Step 7: The handoff sheet

Before the file goes to the printer or the screen, one page records what was made, with what, and who approved it. Both copies end with the same conditions as the source sheet, so the sheet still makes sense when it's passed to someone who never read this article. The owner signs the source sheet and the final; the manager runs the check; the person at the counter is a good second reader for the distance test.

Sample (fictional, nothing generated)

MENU BOARD HANDOFF SHEET - SAMPLE (fictional, Corner Test Kitchen)
Method:            Documentation-based workflow; no account or model test
                   performed for this sample.
Source sheet:      v3 / 2026-10-07, approved by [owner] on [date]
Product / account: [Gemini API or Google AI Studio] / [account or project]
                   Paid or unpaid use: [per the account; see Data below]
Model:             gemini-nano-banana-2.1
Access check:      18+ / available region / business use / age verified:
                   [check in the account - not verified for this sample]
Inputs check:      No sensitive, confidential, personal or private business
                   information in prompts or uploads: [confirm]
Requested size:    4:1 at 2K = 4096 x 1024 px
Received size:     [not generated - sample]
Search grounding:  Off
Assets uploaded:   None (no logo, no photos)
Rights:            Text and shapes only; no third-party assets
SynthID:           Present per Google; not removed or hidden
Text check:        [not run - no draft exists for this sample]
Final text set in: [image as generated / design program as live text]
Required notices:  [none for fictional sample; owner supplies real ones]
Known gaps:        4K cost unresolved in Google docs; AI Studio access and
                   billing not verified; readability not proofed
Print/screen spec: 48 x 12 in; printer minimum resolution: [ask printer]
Proof:             [actual-size print or on-screen proof] checked by ____
Final approval:    ______________________  Date: ____________

CONDITIONS - KEEP WITH EVERY COPY OF THIS SHEET
Status:  Untested. Written from Google's Gemini API docs, read 2026-10-07;
         no image generated, no account or model tried. Every draft is
         unchecked until it passes the full check.
Model:   gemini-nano-banana-2.1 (Gemini Nano Banana 2.1), Gemini API or
         Google AI Studio. Not confirmed for your account or the Gemini app.
Access:  Google's terms: 18 or older, available region, professional or
         business use. Check your account and age verification first.
Inputs:  Nothing sensitive, confidential or personal, and no private
         business information (food costs, suppliers, staff or guest
         details, unannounced changes), on any tier. Your own sheet too.
Data:    Unpaid (e.g. AI Studio without billing): Google uses inputs and
         outputs to improve products; human reviewers may read them.
         Paid (API via a Cloud project with active billing; AI Studio with
         access to such a project, or a Workspace enterprise account): not
         used to improve products, but logged for a limited period - not
         zero retention. EEA, Switzerland, UK: paid data terms apply to all
         use. Search grounding on: prompts and output kept 30 days.
Cost:    Standard Gemini API, pricing page of 2026-10-07: no free tier.
         Image output only $0.0336 (1K) or $0.0504 (2K) per image; input,
         text and thinking tokens extra. Not an AI Studio quote. No 4K
         budget: Google's guide says 2,520 tokens, its pricing page 3,780.
Rules:   Grounding off. Upload only images you have rights to. Keep
         SynthID. Owner approves every word, price and notice. Recheck the
         whole draft after any change. Proof at real size and distance. If
         text keeps failing, set it as editable text in a design program.

Blank copy

MENU BOARD HANDOFF SHEET
Method:            Documentation-based workflow; no account or model test
                   performed unless recorded below.
Source sheet:      version ____ / date ____, approved by ______ on ______
Product / account: ______________________ / ______________________
                   Paid or unpaid use: ________ (see Data below)
Model:             gemini-nano-banana-2.1 (this sheet is written for it)
Access check:      18+ / available region / business use / age verified:
                   ______________________
Inputs check:      No sensitive, confidential, personal or private business
                   information in prompts or uploads: ______________
Requested size:    ratio ____ at ____ = ______ x ______ px
Received size:     ______ x ______ px
Search grounding:  Off / On (should be Off)
Assets uploaded:   ______________________
Rights:            ______________________
SynthID:           Present per Google; not removed or hidden
Text check:        all lines Pass? ____  check table attached? ____
Final text set in: image as generated / design program as live text
Required notices:  ______________________ (owner supplies and approves)
Known gaps:        ______________________
Print/screen spec: ______________________ (from printer or screen vendor)
Proof:             actual-size print / on-screen proof, checked by ______
Final approval:    ______________________  Date: ____________

CONDITIONS - KEEP WITH EVERY COPY OF THIS SHEET
Status:  Untested. Written from Google's Gemini API docs, read 2026-10-07;
         no image generated, no account or model tried. Every draft is
         unchecked until it passes the full check.
Model:   gemini-nano-banana-2.1 (Gemini Nano Banana 2.1), Gemini API or
         Google AI Studio. Not confirmed for your account or the Gemini app.
Access:  Google's terms: 18 or older, available region, professional or
         business use. Check your account and age verification first.
Inputs:  Nothing sensitive, confidential or personal, and no private
         business information (food costs, suppliers, staff or guest
         details, unannounced changes), on any tier. Your own sheet too.
Data:    Unpaid (e.g. AI Studio without billing): Google uses inputs and
         outputs to improve products; human reviewers may read them.
         Paid (API via a Cloud project with active billing; AI Studio with
         access to such a project, or a Workspace enterprise account): not
         used to improve products, but logged for a limited period - not
         zero retention. EEA, Switzerland, UK: paid data terms apply to all
         use. Search grounding on: prompts and output kept 30 days.
Cost:    Standard Gemini API, pricing page of 2026-10-07: no free tier.
         Image output only $0.0336 (1K) or $0.0504 (2K) per image; input,
         text and thinking tokens extra. Not an AI Studio quote. No 4K
         budget: Google's guide says 2,520 tokens, its pricing page 3,780.
Rules:   Grounding off. Upload only images you have rights to. Keep
         SynthID. Owner approves every word, price and notice. Recheck the
         whole draft after any change. Proof at real size and distance. If
         text keeps failing, set it as editable text in a design program.

Is Nano Banana 2.1 the right tool for your board?

• Reasonable to try if you want a few layout ideas for a wide board, your account can use the model or you'll pay the API rate, and someone will check every line.
• Probably not if you need print-ready artwork today with no proof step, if prices change often enough that you'll be regenerating constantly, or if the board carries required notices you can't check line by line.
• Either way, the source sheet, the check table and the handoff sheet work without any AI at all. Hand them to a designer and you get the same control over prices.

More practical guides are in RestaurantLaunch Technology.

Sources

All read October 7, 2026.

• Gemini API release notes, October 6, 2026 entry, Google AI for Developers
• Gemini Nano Banana 2.1 model page, last updated October 6, 2026, Google AI for Developers
• Image generation guide (model selection, features, editing, limitations, aspect ratios and image size), last updated October 6, 2026, Google AI for Developers
• Gemini Developer API pricing, Gemini Nano Banana 2.1 section, last updated October 7, 2026, Google AI for Developers
• Gemini API Additional Terms of Service, last updated April 28, 2026, Google AI for Developers
• Available regions for Google AI Studio and Gemini API, last updated April 28, 2026, Google AI for Developers