How to Shoot Stunning Food Photography with Nano Banana 2

Food photography is one of those disciplines where the gap between “good enough” and “actually great” comes down to lighting, composition, and about three hours of fussing with a lemon wedge. Nano Banana 2 — Google’s viral AI image generator built on Gemini 3.1 Flash Image, launched February 26, 2026 — compresses all of that into a well-crafted prompt. The model renders textures, steam, sauces, and garnishes with a fidelity that’s genuinely unsettling if you’ve ever tried to shoot a burger that doesn’t droop under studio lights.

This tutorial walks you through creating professional-grade food photography with Nano Banana 2, from your first prompt to refined final shots. Whether you’re building a restaurant menu, a recipe blog header, or a social media content library, the prompts and techniques below will get you there without a single prop, stylist, or $4,000 camera body.

What You’ll Achieve

By the end of this guide, you’ll have a repeatable workflow for generating food images that look like they belong in Bon Appétit — hero shots, editorial spreads, close-up texture studies, and social-ready square formats. You’ll also understand how to control lighting, mood, and plating style through prompt language alone, and how to maintain subject consistency across a set of images using Nano Banana 2’s multi-character (up to 5 subjects) consistency feature.

Requirements

You need access to Nano Banana 2. The generator is available through the Gemini app (mobile and web), Google AI Studio, the Gemini API, and Vertex AI for enterprise workflows. The Gemini app is the fastest way to start — sign in with a Google account, open the image generation panel, and you’re generating in under a minute. AI Studio gives you more parameter control and is free for most usage tiers. Vertex AI is where teams with volume needs and compliance requirements live.

Nano Banana Pro unlocks higher-resolution outputs and extended subject consistency sessions — worth it if you’re producing a full menu or brand asset library. The standard tier handles one-off shots and experimentation without issue. All outputs carry SynthID watermarks, Google’s invisible watermarking layer — invisible to the eye, verifiable by detection tools, and present regardless of tier.

The Anatomy of a Great Food Prompt

Food photography prompts that actually work share a consistent structure: subject + plating style + lighting + angle + mood + technical finish. Skip any of these and the model fills in the gaps with its own defaults, which tend toward bland overhead shots with flat light. You want to drive every variable.

Let’s break down the components before hitting the prompts:

Subject: Be specific. “Pasta” is nothing. “Cacio e pepe with visible pepper crust and glossy sauce coating wide spaghetti” is something the model can render with texture.

Plating style: Rustic, fine dining, street food, deconstructed, family-style — each triggers a different visual vocabulary.

Lighting: Side light, backlight, and window light all read differently. “Soft natural window light from the left” beats “good lighting” every time.

Angle: Overhead (flat lay), 45-degree, eye-level, and close-up macro all serve different purposes. State it explicitly.

Mood: Warm and cozy, bright and fresh, dark and moody, minimalist and clean — this shapes color grading and shadow behavior.

Technical finish: Nano Banana 2 outputs at 4K resolution. Prompting for “sharp focus”, “shallow depth of field”, or “cinematic color grade” pushes the model toward photorealistic camera behavior rather than illustrated output.

Ready-to-Use Food Photography Prompts

These are copy-paste ready. Each one is built on the structure above. Swap ingredients, adjust lighting direction, or change the mood phrase to generate variants from the same base.

Hero shot — restaurant-quality main dish:

Close-up food photography of a pan-seared salmon fillet on a white ceramic plate, crispy skin with caramelized edges, garnished with lemon beurre blanc sauce and microgreens, soft natural window light from the left, 45-degree angle, shallow depth of field, warm neutral tones, photorealistic, 4K, fine dining aesthetic

The 45-degree angle is the workhorse of restaurant photography — it shows height, depth, and plating detail simultaneously. Shallow depth of field pulls the eye to the salmon while the plate edge softens into blur. Swap salmon for duck breast, scallops, or a vegetable tart and the structure holds.

Dark and moody editorial — chocolate dessert:

Editorial food photography of a dark chocolate lava cake on a matte black slate plate, molten center oozing, dusted with cocoa powder and gold leaf accent, dramatic side lighting with deep shadows, 45-degree angle, very shallow depth of field, dark moody atmosphere, rich warm tones, photorealistic, 4K, Michelin-star plating style

Dark moody food photography performs extremely well on editorial channels and premium menus. The “matte black slate plate” grounds the color palette. “Gold leaf accent” gives the model a single luxury cue that propagates across the whole image’s tone.

Flat lay — breakfast overhead:

Overhead flat lay food photography of a rustic wooden table with a full breakfast spread: a stack of fluffy buttermilk pancakes with maple syrup dripping down the sides, fresh blueberries, a glass of orange juice, a small white pitcher of cream, scattered wildflowers as props, bright natural daylight, clean white and wood tones, sharp focus throughout, photorealistic, 4K

Flat lays work because composition control is total — nothing competes with the overhead angle for prop arrangement. The “scattered wildflowers” prompt is a small detail that consistently upgrades the lifestyle feel. Remove them and you get a catalog shot. Keep them and you get a Sunday morning.

Social media square — vibrant street food:

Square format food photography for social media, close-up of a Korean fried chicken sandwich on a brioche bun, double-stacked crispy chicken thighs with gochujang glaze, pickled daikon, fresh cilantro, sesame seeds, hand-held with neon bokeh background, bright punchy colors, eye-level angle, shallow depth of field, street food energy, photorealistic, 4K

“Square format” is an explicit instruction Nano Banana 2 respects, saving you a crop. “Neon bokeh background” creates that urban street food visual context without needing to describe an entire environment. “Punchy colors” is deliberately loose — it lets the model interpret the color grading rather than locking it to specific hues, which tends to produce more natural results than over-prescribing.

Product shot — artisan coffee:

Product food photography of a single ceramic cup of cappuccino with perfect latte art rosette, placed on a weathered marble surface with coffee beans scattered nearby, steam rising from the cup, soft backlight creating rim lighting effect, 45-degree angle, shallow depth of field, warm beige and terracotta tones, clean minimal composition, photorealistic, 4K, café lifestyle aesthetic

“Steam rising” is one of Nano Banana 2’s party tricks — the model renders vapor convincingly without it looking like a CGI explosion. Backlight plus steam is a classic combination that the prompt exploits deliberately. “Weathered marble surface” does more compositional work than you’d expect: it sets the café aesthetic without describing a whole room.

Consistency shoot — three-course menu set:

Food photography series, consistent style across all dishes: [Dish 1: burrata with heirloom tomatoes and basil oil, white marble surface], [Dish 2: seared duck breast with cherry jus and wilted greens, same white marble surface], [Dish 3: panna cotta with raspberry coulis and edible flowers, same white marble surface]. Soft diffused natural light, 45-degree angle, shallow depth of field, fine dining plating, warm neutral tones, photorealistic, 4K

This is Nano Banana 2’s subject consistency feature applied to a full menu shoot. The bracketed structure signals three distinct subjects while the shared surface, lighting, and angle instructions maintain visual cohesion. You get a matched set that looks like it came from the same two-hour studio session. The model handles up to five subjects in a consistency session — enough for a full tasting menu.

Texture macro — artisan bread:

Extreme close-up macro food photography of a freshly baked sourdough loaf, cross-section showing open crumb structure with irregular air pockets, crust scoring pattern visible, flour dusted surface, warm side lighting raking across the surface to emphasize texture, 90-degree angle, razor-sharp focus on crumb detail with blurred background, earthy warm tones, photorealistic, 4K

“Raking light” — light that skims across a surface at a low angle — is the standard technique for making textures pop in real photography. Prompting for it works because the model has seen thousands of food photography examples where this technique appears. “Razor-sharp focus” paired with “blurred background” explicitly sets bokeh behavior for macro shots.

Recipe blog hero — seasonal salad:

Food photography hero image for a recipe blog, wide composition of a vibrant summer salad in a large ceramic bowl, mixed greens, watermelon cubes, feta cheese, mint leaves, toasted pine nuts, light lemon dressing drizzled mid-pour from a small pitcher, bright airy atmosphere, soft natural light, slight 30-degree angle, sharp focus on center, clean white linen surface, fresh and appetizing mood, photorealistic, 4K

“Mid-pour” is a motion cue that adds life to a static shot. The model renders it as a frozen action moment — dressing suspended in the air — which is exactly what food photographers achieve with high shutter speeds. “White linen surface” signals the blog-friendly bright aesthetic that performs on Pinterest and recipe sites.

Tips, Tricks, and Things That Actually Matter

Pro tip ✅

Always specify the surface material in your prompt. “White marble”, “weathered oak wood”, “dark slate”, “rustic terracotta tile” — the surface sets the entire color palette and texture story of the image. Skip it and the model picks something neutral and forgettable.

Pro tip ✅

For consistent menu shoots, build a style string and paste it at the end of every prompt in the set: “soft diffused natural light, 45-degree angle, shallow depth of field, fine dining plating, warm neutral tones, photorealistic, 4K, white marble surface.” Treat it like a Lightroom preset — apply it to every shot in the series.

Warning ⚠️

All Nano Banana 2 outputs carry SynthID watermarks — Google’s invisible digital watermarking layer. They’re not visible in the image and won’t affect commercial use in most contexts, but if you’re delivering assets to a client, disclose this. Detection tools can identify SynthID-marked images, and some platforms are starting to flag AI-generated content in metadata.

Pro tip ✅

If your first generation is close but the lighting is wrong, don’t regenerate from scratch. Use the edit workflow in the Gemini app to target the lighting specifically: “adjust the lighting to be softer and more diffused, reduce harsh shadows on the left side.” Nano Banana 2’s editing workflow preserves the rest of the composition while making targeted changes — far more efficient than rewriting the full prompt.

Note 💡

Nano Banana 2 handles text rendering accurately — a direct upgrade over earlier image generators that mangled words into abstract decoration. If you need a menu card, a chalkboard special, or a price tag in the shot, include the exact text in quotes in your prompt: “chalkboard in background reading ‘Today’s Special: Burrata $18′”. Nail that and Midjourney V7 starts looking nervous.

Pro tip ✅

Access via AI Studio if you want to run batches. The API interface lets you loop through ingredient variations — useful for generating a full menu’s worth of shots in one session. Vertex AI is the move for restaurant groups or agencies that need usage logging, enterprise SLAs, and higher rate limits.

Avoid 🚫

Don’t use generic mood words like “beautiful” or “delicious” — the model has seen them paired with everything from gas station hot dogs to wedding cakes. They add zero signal. Replace them with specific descriptors: “appetizing with visible steam”, “glossy sauce with light catching the surface”, “crispy edges with visible browning texture”. Specific sensory language translates directly into visual output.

What Nano Banana 2 Gets Right (and Where to Push It)

The model’s strength in food photography comes down to two things: texture rendering and lighting physics. Crispy surfaces look crispy. Wet sauces catch light like wet sauces should. Steam behaves like steam. These aren’t easy problems — earlier image generators routinely produced food that looked like it was coated in a uniform plastic sheen regardless of what the prompt said.

Where you’ll still need to push is in very specific cultural or regional food presentations. If you’re prompting for a traditional Japanese kaiseki course with highly specific plating rules, or a West African dish with particular regional presentation conventions, be explicit and detailed — the model’s defaults will drift toward familiar Western fine dining aesthetics unless you steer it firmly. Real-time web grounding helps here: Nano Banana 2 can pull visual references from current web content, so specifying a recently trending food style or a specific restaurant’s plating aesthetic gets you closer to that reference point than you might expect.

Compared to Nano Banana (the original), the jump in consistency across multi-image sets is the biggest practical improvement for food photographers building full menus or content libraries. The original required significant prompt engineering to keep lighting and surface materials matched across shots. Nano Banana 2 handles it with a single style string — which is the kind of quality-of-life improvement that sounds small until you’re generating your fortieth dish and everything still matches.

Start Plating

Food photography is one of the highest-ROI use cases for Nano Banana 2 — the visual language is specific enough that precise prompts produce precise results, and the commercial applications are immediate. A restaurant owner can generate a full menu’s imagery in an afternoon. A food blogger can build a month of header images before lunch. An agency can deliver a client brief without scheduling a single studio day.

Copy the prompts above, swap in your dishes, and start with the surface material — everything else builds from there. The lemon wedge is no longer your problem.

promptyze

See Full Bio