How to Create Marketing Mockups with Perfect Text in Nano Banana 2

For years, AI image generators treated text like a decoration — something to smear across a composition and hope nobody looked too closely. Blurry letters, missing serifs, words that spelled something vaguely adjacent to what you asked for. Then Google launched Nano Banana 2 (powered by Gemini 3.1 Flash Image) on February 26, 2026, and suddenly precise text rendering is table stakes. Marketing teams, freelance designers, and solo founders are already using it to generate client-ready mockups in minutes — no Photoshop, no placeholder text, no apologies.

This tutorial walks you through building marketing mockups with Nano Banana 2 from scratch: product shots, social ads, packaging, billboards, and more. Every prompt below is copy-paste ready. If you’ve been burned by garbled AI text before, this is the one to revisit.

What You’ll Achieve

By the end of this guide, you’ll know how to generate polished marketing mockups — product labels, social media ads, storefront signage, and promotional banners — with legible, accurate text baked directly into the image. You’ll also know which prompt structures consistently produce clean typography, how to maintain subject consistency across a series of mockups, and how to get 4K-quality output suitable for client presentations.

What You Need

Nano Banana 2 runs on Gemini 3.1 Flash Image and is available through four channels: the Gemini app (the easiest entry point), Google AI Studio (free, great for experimenting with parameters), the Gemini API (for building it into your own tools), and Vertex AI (for enterprise workflows). For this tutorial, AI Studio works perfectly — it’s free, gives you direct access to the model, and shows you exactly what you’re sending. No subscription required to start.

The Golden Rule: Tell It the Text Exactly

Nano Banana 2’s text rendering is precise, but it’s not psychic. The single most important habit you can build is quoting your text explicitly inside the prompt — in quotation marks, spelled out exactly as you want it to appear. Don’t write “add a tagline.” Write: tagline reads “The future tastes better.” That distinction alone separates clean mockups from garbled ones.

Pro tip ✅

Always wrap your intended text in quotation marks inside the prompt. Nano Banana 2 treats quoted strings as literal rendering targets, which dramatically improves accuracy on brand names, slogans, and product copy.

Prompt 1: Product Packaging Label

Product packaging label for a cold brew coffee bottle. Clean minimal design, matte black label, white sans-serif font. Label reads "DARKWAKE" in large bold letters, subtitle reads "Cold Brew Concentrate", bottom text reads "12 FL OZ". Dark studio background, 4K, photorealistic product photography style.

This prompt hits all the structural notes: it specifies the product type, the surface the text appears on, the font style, and every line of copy in explicit quotes. The “4K, photorealistic product photography style” instruction pushes Nano Banana 2 toward output that looks like it came from an actual product shoot rather than a render. If you want a variant with a kraft paper label and earthy tones instead, swap “matte black label, white sans-serif font” for “kraft paper label, dark brown serif font” — the rest of the prompt stays identical.

Prompt 2: Social Media Ad — Square Format

Instagram square ad mockup, 1:1 ratio. Lifestyle photo of a woman drinking from a reusable water bottle, golden hour outdoor lighting. Overlay text in bold white font reads "STAY HYDRATED" at top, smaller text reads "Shop now at drinkpure.com" at bottom. Clean minimal layout, slight dark gradient at bottom for text legibility. 4K resolution.

Social ads live or die on text legibility, and this prompt handles it by specifying a dark gradient under the bottom copy — a real designer trick that Nano Banana 2 understands and executes. The explicit URL in the prompt (“drinkpure.com”) tests the model’s ability to render domain-style text accurately. You can swap the subject entirely — change the lifestyle photo to a product flat lay — without touching the text structure, which is the point of working with consistent prompt architecture.

Pro tip ✅

For social ad mockups, always specify the aspect ratio and mention a gradient or background treatment beneath your text. Nano Banana 2 generates the text and the treatment together, so you get legible copy without needing to composite anything in post.

Prompt 3: Storefront Window Signage

Exterior view of a modern café storefront, daytime, natural light. Large vinyl window decal in white reads "GROVE COFFEE" in uppercase serif letters. Below it, smaller text reads "Open Daily · 7am–8pm". Realistic window reflection, passersby slightly blurred in background, editorial photography style, 4K.

Storefront mockups are a staple of brand presentations — clients want to see their identity in the wild before they commit to a printer. This prompt uses “editorial photography style” to push the output toward something that looks candid and real rather than staged. The operating hours (“Open Daily · 7am–8pm”) are a good stress test for Nano Banana 2’s text handling: special characters, mixed formats, real-world punctuation. If you need a night version, add “neon-lit evening scene, warm interior light spilling through window” and drop “daytime, natural light.”

Prompt 4: Promotional Banner — Horizontal

Horizontal promotional web banner, 16:9 ratio. Bold geometric background in deep navy and gold. Centered headline text reads "SUMMER SALE" in large gold serif font. Subheading reads "Up to 50% off sitewide". Call-to-action button reads "SHOP THE SALE". Clean luxury brand aesthetic, high contrast, 4K.

Web banners are where a lot of AI image generators collapse — too many text elements, inconsistent sizing, letters bleeding into the background. Specifying high contrast and naming each text element separately (headline, subheading, CTA button) gives Nano Banana 2 a clear hierarchy to work with. The “luxury brand aesthetic” instruction keeps the output from going garish with the navy-and-gold palette.

Note 💡

Nano Banana 2 supports subject consistency across up to five characters or elements in a sequence. If you’re building a multi-asset campaign, describe your product or mascot in detail in the first prompt, then reference that same description in follow-up prompts. The model carries visual consistency across the series without you needing to re-upload a reference image every time.

Prompt 5: Product Ad — Portrait Format (Mobile)

Vertical portrait ad mockup, 9:16 ratio, for a skincare serum. Minimal white background, product bottle centered. Elegant thin serif font overlay reads "LUMIÈRE SÉRUM" at top. Below product, text reads "Clinically tested. 28-day results." Small footnote text at bottom reads "*Results may vary. Fragrance-free formula." Soft studio lighting, premium beauty brand aesthetic, 4K.

This prompt is deliberately ambitious — it asks for a headline, a body line, and fine-print footnote text, all in the same frame. Nano Banana 2’s text rendering handles multi-tier copy better than any previous generation of the model, but the footnote is where you’ll want to check your output carefully. If the small text comes out blurry, add “footnote text crisp and legible, high resolution fine print” to the prompt and regenerate.

Warning ⚠️

The smaller the text in your mockup, the more likely Nano Banana 2 is to introduce subtle rendering errors — a transposed letter, a missing period. Always zoom into fine print in your output before sending it to a client. This isn’t a bug unique to Nano Banana 2; it’s physics: small text at any resolution is harder to pin down. Double-check everything below roughly 12pt equivalent.

Prompt 6: Event Poster

Music festival poster, portrait orientation. Dark moody background with abstract smoke and purple/orange gradient lighting. Main headline reads "NOCTURNE FEST 2026" in bold condensed white font. Subtext reads "July 18–20 · Austin, TX". Below that, smaller text reads "Featuring: Avalanche Sound, Luna Kross, Desert Static". Bottom footer reads "Tickets at nocturnefest.com". Gritty editorial poster aesthetic, high contrast, 4K.

Event posters pack more text than almost any other marketing format, which makes them a genuine benchmark for text-rendering quality. Listing the performers explicitly in the prompt (“Avalanche Sound, Luna Kross, Desert Static”) tests whether Nano Banana 2 preserves invented proper nouns accurately — and it mostly does. If one act’s name comes out garbled, isolate it in a follow-up edit prompt rather than regenerating the whole image.

Prompt 7: Packaging — Side Panel / Ingredient List Style

Close-up of a cereal box side panel, warm morning kitchen background, slightly angled. Printed label section reads "INGREDIENTS: Whole grain oats, brown sugar, almonds, sea salt, natural flavors." Below that, text reads "NET WT 12 OZ (340g)". Clean consumer packaged goods typography, realistic print texture on cardboard, 4K product photography.

This one’s a curveball — ingredient panels are dense, boring, and exactly the kind of copy that AI image generators used to turn into alphabet soup. Nano Banana 2 handles it because the prompt specifies “consumer packaged goods typography” and “realistic print texture,” which anchors the model to a known visual style with predictable text layout conventions. Use this prompt structure any time you need to mockup the back or side of packaging, not just the hero face.

Pro tip ✅

When prompting for packaging mockups, reference the printing substrate explicitly — cardboard, glass, matte plastic, frosted film. Nano Banana 2 adjusts text rendering to match the surface, so “serif font on frosted glass” will look genuinely different from “serif font on kraft paper.” This also makes your mockups more convincing to clients who know their materials.

Prompt 8: Billboard Mockup — Outdoor / OOH

Realistic outdoor billboard mockup on a city street, overcast daytime light. Billboard reads "YOUR CITY. YOUR COFFEE." in massive bold white sans-serif letters on a dark charcoal background. Small logo area bottom-right reads "GROVE COFFEE". Urban street scene, slight lens distortion, editorial photography style, 4K.

Out-of-home mockups close a lot of pitches. Showing a client their brand on a billboard — even an AI-generated one — makes the abstract concrete. The “slight lens distortion” instruction makes the image feel like it was shot from the street rather than rendered from a template. Pair this with the storefront prompt from earlier (using the same brand name “GROVE COFFEE”) and you have a two-asset location shoot without leaving your desk.

Pro tip ✅

Nano Banana 2 outputs images with SynthID watermarks embedded invisibly in the pixel data. They won’t appear visually in your mockup and won’t affect client presentations, but they’re there — and they identify the image as AI-generated. For any mockup going into a real ad campaign, your production team will replace the AI asset with a properly licensed version anyway. Treat Nano Banana 2 output as a presentation layer, not a final deliverable.

Working the Edit Loop

Nano Banana 2’s editing workflow lets you refine an existing generated image with text instructions rather than regenerating from scratch. Once you have a base mockup you like, you can send it back to the model with targeted edits: “change the headline color from white to gold,” “make the product label larger and centered,” “darken the background gradient.” This is significantly faster than iterating on the full prompt and avoids losing elements you already like. In AI Studio, the edit-in-context interface makes this loop straightforward — generate, inspect, describe the change, regenerate the patch.

Note 💡

If you’re accessing Nano Banana 2 through the Gemini API directly, the image editing endpoint accepts both the original image and a delta prompt as inputs. For programmatic mockup workflows — generating dozens of size variants of the same ad — this is far more efficient than calling the generation endpoint fresh each time.

Why This Changes the Mockup Workflow

Before Nano Banana 2, the standard mockup process went: rough prompt → broken text → Photoshop → fix text manually → client presentation. The AI was a starting point, not an endpoint. With accurate text rendering and 4K output, you can skip the Photoshop pass entirely for concept-stage work. That’s not a small thing — for freelancers billing hourly, it’s the difference between two hours and twenty minutes per deliverable. For in-house teams running fast-cycle campaigns, it means same-day concept reviews instead of next-morning turnarounds. The prompts above aren’t exhaustive; they’re a framework. Once you internalize the structure — quote your text, specify hierarchy, anchor to a visual style, call out the resolution — you can adapt it to any format a client throws at you.

promptyze

See Full Bio