Skip to content
Nano Banana

How to Use Nano Banana 2 to Generate Presentation Slides and Visuals

Learn how to use Nano Banana 2 to generate 4K slide backgrounds, consistent characters, charts, and icons for presentations — with copy-paste prompts.

9 min read
How to Use Nano Banana 2 to Generate Presentation Slides and Visuals

Presentation design has always been the part of any project where ambition meets reality. You have a great idea, a deadline, and absolutely zero desire to spend four hours in Canva hunting for stock photos that don’t look like stock photos. Nano Banana 2 — Google’s image generator built on Gemini 3.1 Flash Image, launched February 26, 2026 — cuts most of that friction. It renders precise text into images, keeps characters consistent across multiple slides, outputs at 4K resolution, and pulls real-world context via live web grounding. That combination makes it genuinely useful for building presentation visuals, not just pretty pictures.

This tutorial covers the full workflow: from spinning up Nano Banana 2 in the Gemini app (or via API) to writing prompts that produce slide-ready assets. You’ll get copy-paste prompts, settings to tweak, and the protips that separate decent results from ones you’d actually put in front of a client.

What You’ll Achieve

By the end of this guide you’ll be able to generate title slide backgrounds, data visualization mockups, icon sets, character-consistent presenter avatars across multiple slides, and editorial-style hero images — all at 4K — with readable text baked right into the image. That last part alone is worth paying attention to; most AI image generators still treat text as decorative squiggles.

Requirements

You need access to Nano Banana 2 through one of four routes. The Gemini app (gemini.google.com) is the fastest for one-off generation. AI Studio (aistudio.google.com) is better for iterating quickly with prompt history. The Gemini API gives you programmatic access for batch generation. Vertex AI is the enterprise path if you’re building this into a production pipeline. A Google account covers the first two. API and Vertex access require a Google Cloud project with billing enabled. SynthID watermarks are embedded in all outputs by default — they’re invisible to the human eye but detectable by Google’s verification tools, so factor that in if you’re publishing commercially.

Step 1 — Set Your Output Specs Before You Touch a Prompt

Before writing a single prompt, decide on aspect ratio. Standard presentation slides are 16:9. A square image dropped into a widescreen deck looks like a design accident. In the Gemini app, use the aspect ratio selector in the generation panel. In AI Studio or via API, pass the parameter directly in your request. For 4K output specify resolution: 4K or the equivalent pixel dimensions (3840×2160 for landscape, 2160×3840 for portrait) in your API call. Higher resolution means slightly longer generation time but you get assets you can scale without blur — relevant if slides end up printed or displayed on a large screen.

Pro tip ✅

Always generate at 4K even if your deck is 1080p. Downscaling a 4K image to fit a slide preserves sharpness in ways that native 1080p generation often doesn’t. Think of it as headroom.

Step 2 — Generate a Title Slide Background

Title slides set the visual tone for everything that follows. You want something bold but not distracting — the text you overlay later needs to stay legible. The trick is to prompt for images with deliberate negative space or subtle compositional anchors.

Wide cinematic background for a tech company presentation title slide. Deep navy blue gradient, abstract flowing light trails in electric blue and white, subtle geometric grid overlay. Left third of image intentionally dark and empty for white headline text. 4K resolution, 16:9 landscape, no text in image.

That “intentionally dark and empty for white headline text” instruction is doing real work here. Without it, Nano Banana 2 will fill the frame. The phrase guides the composition so you have somewhere to put your actual slide title.

Minimalist title slide background for a sustainability report presentation. Soft olive green and warm sand gradient. Top half features abstract aerial view of a green canopy, slightly out of focus. Bottom third clean and light for dark overlay text. 16:9, 4K, no text, no logos.

Step 3 — Generate Section Divider Slides

Multi-section decks need visual breaks that reset the audience’s attention without requiring a new design system. Section dividers should feel like part of the same family as the title slide — same palette, different energy.

Section divider slide background for a financial services presentation. Dark charcoal background, single bold diagonal stripe in gold, subtle texture like brushed metal. Strong contrast, professional, no text, no UI elements, 16:9, 4K.

Generate three or four of these with slight palette variations and you have a coherent slide library in under ten minutes.

Pro tip ✅

Use Nano Banana 2’s subject consistency feature when generating a series of section dividers. Set the same “anchor” subject — a geometric shape, a character, a motif — and it will appear consistently across all generated images. This is what makes a deck look designed rather than assembled from stock photos.

Step 4 — Generate a Consistent Presenter Avatar Across Multiple Slides

Subject consistency across up to five characters is one of Nano Banana 2’s more useful features for presentation work. If your deck features a fictional spokesperson, a branded mascot, or an illustrated persona, you can keep that character recognizable from slide to slide without redrawing from scratch.

Professional illustrated portrait of a woman in her 40s, short natural gray hair, warm brown skin, wearing a navy blazer, confident direct expression, soft studio lighting, white background, editorial illustration style, clean lines. Consistent character seed: "Presenter_01".

Generate this once, then reference the character seed in follow-up prompts to keep her face, hair, and clothing consistent as she appears in different contexts — at a podium, next to a chart, gesturing toward text. The consistency feature means you’re not introducing a new character on slide 14 who looks like a different person entirely.

Same illustrated woman from character seed "Presenter_01", now shown from the waist up, standing beside a large monitor showing bar charts, pointing toward the screen, confident posture. Same navy blazer, gray hair, warm brown skin. Soft office background, editorial illustration style, 4K.

Warning ⚠️

Character consistency works best when you describe the anchor traits (hair, skin tone, clothing color, art style) explicitly in every follow-up prompt. Don’t rely on the seed alone. Think of the seed as a shortcut, not a guarantee.

Step 5 — Generate Data Visualization Mockups

Nano Banana 2’s precise text rendering makes it viable for generating chart-style visuals — not as a replacement for actual data tools, but as mockup assets or decorative infographic elements for slides that communicate direction rather than exact figures.

Clean infographic-style bar chart showing fictional revenue growth over five years. Bars in teal and coral on a white background. X-axis labeled "2020, 2021, 2022, 2023, 2024". Y-axis labeled in thousands. Title text: "Revenue Growth". Numbers above each bar. Minimal design, presentation-ready, no grid lines, 16:9, 4K.

The key is explicitly including the label text you want rendered in the prompt. Nano Banana 2’s text engine will attempt to render those strings. Verify the output — text rendering at this level is impressive but not infallible, and you may need to regenerate once or twice to get clean results on complex labels.

Minimalist pie chart with three segments: 45% in deep blue labeled "North America", 30% in teal labeled "Europe", 25% in coral labeled "Asia Pacific". White background, clean sans-serif labels, presentation style, no drop shadows, 4K, 16:9 crop with space for a slide title at top.

Note 💡

Real-time web grounding in Nano Banana 2 means you can reference current visual styles or design trends and the model will reflect them in output. Ask for “2026 minimal presentation aesthetic” and it will have some idea what that actually looks like, rather than serving up 2019 flat design conventions.

Step 6 — Generate Icon Sets and Supporting Visual Assets

Individual slide elements — icons, decorative dividers, small illustrations — often take more time to source than the slides themselves. Generating a consistent icon set in one session is faster than hunting through Noun Project for a matching family.

Set of six flat icons on white background: laptop, speech bubble, bar chart, lightbulb, shield, globe. All icons in deep indigo, consistent line weight, simple outline style, even spacing, no shadows, no text labels, square format, 4K.

Ask for multiple icons in one prompt and Nano Banana 2 will arrange them in a grid. You then crop individual icons in any design tool. It’s not as precise as an SVG icon library, but for a one-off presentation that ships tomorrow, it works.

Single flat icon: a handshake symbol. Deep indigo, consistent with a minimal outline icon set, white background, centered, square format, no text, 4K, clean edges for easy cropping.

Step 7 — Generate Social Media Variants of Your Slide Visuals

Once your slide assets exist, repurposing them for LinkedIn announcements, Twitter/X headers, or Instagram carousels is just a prompt adjustment. Swap the aspect ratio and update the composition brief.

LinkedIn announcement graphic. Bold headline text: "Q1 Results Are In". Dark navy background, electric blue accent stripe on left, clean sans-serif typography, professional tone. 1200x627px crop, no additional text elements, 4K base resolution, space for a logo in bottom right corner.
Instagram carousel slide, square format. Topic: sustainable packaging. Warm cream background, botanical illustration of leaves framing the edges, centered dark text area reading "Did You Know?". Minimal, editorial, magazine-quality, 1:1 aspect ratio, 4K.

Pro tip ✅

Generate the 16:9 presentation version and the 1:1 social version in the same session, referencing the same color palette and typography brief. Nano Banana 2 will keep the visual language consistent across both formats, which means your deck and your social posts actually look like they belong to the same campaign.

Step 8 — Batch Generation via API for Larger Decks

For a 30-slide deck, clicking through the Gemini app one image at a time gets old fast. The Gemini API lets you script batch generation — loop through a list of prompts, write outputs to a folder, then import the whole set into your design tool. AI Studio is useful for testing prompts before committing them to a script. Vertex AI handles this at scale with enterprise-grade rate limits and audit logging, relevant if your organization has compliance requirements around generated assets.

Pro tip ✅

Structure your batch prompts as a template with one variable (the slide topic) and consistent boilerplate (resolution, style, palette). It keeps the visual family coherent and makes the script trivial to maintain. Something like: “[TOPIC], [consistent style string], deep navy and gold palette, 16:9, 4K, no text unless specified.”

Avoid 🚫

Don’t generate text-heavy slides entirely in Nano Banana 2 and use them as final output without review. The text rendering is good — better than anything before it — but it still makes occasional character-level errors on long strings. Treat AI-generated text in images as a draft layer, verify it in your design tool before publishing.

Why This Actually Works for Presentations

Most AI image generators fail at presentations because they produce images that look great in isolation and terrible on a slide — wrong aspect ratio, text that reads as decoration, no consistent visual thread from one image to the next. Nano Banana 2 addresses all three. The 16:9 output fits straight into PowerPoint, Keynote, or Google Slides without cropping. The text rendering means labels and callouts are readable. The subject consistency feature — up to five characters — keeps your visual identity coherent across a 20-slide deck. SynthID watermarks are embedded but won’t show up on a projected slide; for internal presentations that’s a non-issue, for commercial publishing it’s worth checking your organization’s policy.

The real argument for this workflow is time. A competent designer can build a deck in two days. Nano Banana 2 gets you to a first draft of every slide’s visual layer in two hours. What you do with the remaining time is up to you — but “refining” beats “starting from scratch” every single time.

author avatar
promptyze

promptyze

ADMINISTRATOR