Nano Banana 2 Fashion Lookbook: Prompts and Workflow

Fashion photography is expensive, time-consuming, and requires about forty people to agree on a mood board. Nano Banana 2 — Promptyze’s name for the Gemini 3.1 Flash Image generator — cuts that process down to a single afternoon and a well-crafted prompt. Since its launch in late February 2026, the model’s subject consistency feature (holding up to five characters across multiple shots) and 4K output have made it genuinely useful for lookbook creation, not just one-off concept art.

Whether you’re a fashion brand prototyping a seasonal campaign, an indie designer who can’t afford a studio shoot, or a creative director who wants to present mood concepts before committing to a budget, this tutorial walks through the exact workflow. Prompts included. No guesswork.

This guide covers Nano Banana 2 via the Gemini app, AI Studio, and the Gemini API — with notes on where each access point makes the most sense for fashion work.

What You’ll Achieve

By the end of this tutorial, you’ll have a complete AI-generated fashion lookbook: consistent models across multiple shots, varied styling scenarios (editorial, street, product detail, social media crop), precise text overlays for brand names or campaign slogans, and a repeatable workflow you can run for every new collection drop.

What You Need Before You Start

Access to Nano Banana 2 comes through a few doors. The Gemini app (gemini.google.com) is the fastest way to start — log in with a Google account and you’re generating within seconds. For batch work and prompt iteration, AI Studio (aistudio.google.com) gives you a proper playground with parameter controls. If you’re building this into a production pipeline, the Gemini API via Vertex AI or Antigravity handles programmatic generation at scale. For a lookbook prototype, the Gemini app or AI Studio is all you need.

One thing to know upfront: every image Nano Banana 2 generates carries a SynthID watermark — Google’s invisible digital signature embedded in the pixel data. It survives screenshots and edits. For internal mockups and client presentations, this is a non-issue. For final commercial use, factor in your post-production workflow accordingly.

Building Your Subject: The Consistency Anchor

The feature that makes lookbook work actually viable is subject consistency across up to five characters. The trick is establishing your model’s visual identity in the first prompt with enough specificity that the model can maintain it across shots. Think of it as a character sheet: physical features, skin tone, hair, and a defining style element.

Here’s a strong character-definition prompt to open any lookbook session:

Portrait photo, fashion editorial. A woman in her late 20s, sharp angular jawline, deep brown skin, cropped natural hair, strong brow. Wearing a structured cream wool blazer with exaggerated shoulders. Neutral expression, direct gaze into camera. Studio lighting, soft diffusion, white seamless backdrop. 4K, high detail, photorealistic.

This prompt does three things well: it locks in physical traits the model can reference in subsequent shots, it sets a photorealistic tone that carries through the lookbook, and it specifies the lighting environment so follow-up shots feel cohesive. The “4K, high detail” instruction at the end consistently pushes Nano Banana 2 toward sharper texture rendering — particularly visible in fabric detail.

Pro tip ✅

Write your character description once and save it as a text snippet. Paste it at the start of every new prompt in the session. Subject consistency in Nano Banana 2 improves significantly when the physical descriptor is explicit in each generation, not just implied from session context.

Lookbook Prompt Set: Six Scenarios, Copy-Paste Ready

Below are six production-ready prompts covering the core scenarios a fashion lookbook needs. Each is built to work with the character anchor above — prepend your character description to any of them for cross-shot consistency.

Scenario 1 — Editorial Full-Length

Full-length fashion editorial photograph. [CHARACTER DESCRIPTION]. Wearing a floor-length camel trench coat, belt cinched at waist, over a black turtleneck and tailored black trousers. Minimal gold jewelry. Standing on a rain-wet urban sidewalk at dusk, bokeh city lights behind her. Shot from a low angle, 35mm lens perspective. 4K, film grain, high contrast.

The low angle and 35mm lens instruction shapes the compositional feel — Nano Banana 2 responds well to camera-style direction and produces distinctly different framing when you specify lens focal length. Film grain gives the output a printed-magazine quality rather than a clinical CGI look.

Scenario 2 — Street Style Candid

Street style photography, candid aesthetic. [CHARACTER DESCRIPTION]. Wearing an oversized vintage denim jacket, wide-leg cargo pants in olive green, white low-top sneakers. Large sunglasses. Mid-stride on a cobblestone street, Paris, morning light, long shadows. Slightly overexposed, natural color palette. 4K, editorial photography.

The “mid-stride, candid aesthetic” instruction shifts the model away from static posed shots. Specify a real city and time of day — Nano Banana 2’s real-time web grounding means it has current visual references for locations, which improves environmental authenticity.

Scenario 3 — Product Detail Close-Up

Close-up product detail photograph, fashion editorial. Woman's hands and wrist area only. Wearing a structured black leather glove on the left hand. On the right wrist, a chunky silver architectural cuff bracelet. Deep navy background, single hard directional light from left. 4K, macro texture detail, no motion blur.

Lookbooks need detail shots, and Nano Banana 2 handles texture rendering — leather grain, metal surface, fabric weave — with strong fidelity when you ask for it explicitly. “No motion blur” prevents the model from adding artistic movement effects that would obscure product detail.

Scenario 4 — Campaign Social Media Crop (Square Format)

Square format fashion campaign image, social media. [CHARACTER DESCRIPTION]. Wearing a bright cobalt blue structured mini dress, architectural neckline. Seated on a white marble staircase, looking upward and off-camera. Bold color contrast: cobalt against white marble. Clean minimalist composition, centered subject, equal negative space above and below. Bright even lighting. 4K, commercial fashion photography.

Specifying square format and equal negative space tells Nano Banana 2 to compose for Instagram rather than a horizontal editorial spread. The “centered subject, equal negative space” instruction is the key — without it, the model defaults to editorial asymmetric framing that crops badly at 1:1.

Scenario 5 — Campaign Text Overlay (Precise Text Rendering)

Fashion campaign photograph with text. [CHARACTER DESCRIPTION]. Wearing a minimalist white linen suit, no accessories. Standing against a solid warm sand-colored background, arms loosely crossed, slight smile. Leave clear visual space in upper third of image for text. Bold sans-serif text overlay at top reading: "SPRING 2026". Smaller text below reading: "NEW COLLECTION". Clean commercial aesthetic. 4K.

Nano Banana 2’s precise text rendering is one of its most practically useful features for campaign work. Short, clean strings — collection names, slogans, dates — render accurately. Keep text instructions short (under six words per line), use plain sans-serif descriptions, and specify position relative to the composition. Attempting long body copy or decorative script fonts produces less reliable results.

Warning ⚠️

Nano Banana 2’s text rendering is solid for short display copy (2-6 words), but don’t push it toward long sentences or mixed-case decorative fonts. If you need precise typographic control for final deliverables, generate the image without text and add type in Figma or Photoshop afterward.

Scenario 6 — Multi-Character Group Shot (Testing the 5-Character Limit)

Fashion editorial group photograph. Three models standing together in a minimalist white studio. Model 1: tall woman, late 20s, deep brown skin, cropped natural hair, wearing a structured cream blazer. Model 2: shorter woman, early 30s, pale skin, long straight red hair, wearing an oversized black coat. Model 3: androgynous figure, mid 20s, medium brown skin, shaved head, wearing a silver puffer vest over a black turtleneck. All three facing camera, slight spacing between them. Flat even studio lighting. 4K, fashion editorial.

Nano Banana 2 handles multi-character consistency up to five subjects. For group shots, describe each character explicitly in the prompt rather than referencing prior session generations — this produces the most stable results. The cleaner and more distinct each character description is (height, skin tone, hair, one key garment), the better the model differentiates between them.

Pro tip ✅

When generating a series of shots with the same character, run 3-4 generations per prompt and pick the best. Nano Banana 2 has natural variation across outputs, and the second or third generation in a batch often has better fabric texture or facial expression than the first.

Editing Workflow: From Generation to Deliverable

Raw Nano Banana 2 output is rarely the finished product for professional lookbooks. Here’s a practical post-generation workflow that designers are using in 2026.

Generate at 4K and download the full-resolution file. Nano Banana 2’s 4K output gives you enough pixel density to crop aggressively for different aspect ratios — a single generation can yield a full-page editorial crop, a square social post, and a banner crop without quality loss. From there, bring the image into your editor of choice: Lightroom for color grading (Nano Banana 2 responds well to subtle dehaze and clarity adjustments), Photoshop for any retouching or composite work, or Figma if you’re building a digital lookbook layout directly.

For brand presentations, watermark removal isn’t necessary — SynthID is invisible to the eye and doesn’t appear in print. What matters at the presentation stage is color consistency across your generated images. If you find your shots drifting in warmth or exposure across prompts, add a consistent lighting descriptor to every prompt: “daylight balanced, 5500K, soft fill light” will keep your lookbook cohesive even across separate generation sessions.

Pro tip ✅

Run all your lookbook prompts in a single AI Studio session with a consistent system prompt that defines your visual style — lighting, color temperature, and film stock aesthetic. Nano Banana 2 in AI Studio gives you a system prompt field that applies to every generation in the session, which saves you from appending the same style descriptors to every individual prompt.

Note 💡

If you’re building a high-volume lookbook pipeline — say, generating 50+ images per collection — the Gemini API via Vertex AI is the right tool. It supports programmatic batch generation, lets you store character descriptions as reusable prompt templates, and handles rate limits more gracefully than the Gemini app’s UI. Antigravity provides a wrapper that simplifies the API authentication flow if you’re not a backend developer by trade.

Prompt Variants: Changing One Parameter, Different Outcome

The fastest way to build prompt intuition with Nano Banana 2 is to run the same base prompt with a single variable changed. Here are three quick variations that demonstrate how much a single parameter shift moves the output:

BASE: Fashion editorial. Woman in structured cream blazer, studio, soft diffusion light, white backdrop. 4K photorealistic.

VARIANT A (change lighting): Same as above but replace "soft diffusion light" with "harsh single-source side lighting, deep shadows."

VARIANT B (change setting): Same as base but replace "white seamless backdrop" with "abandoned industrial warehouse, broken windows, late afternoon sun."

VARIANT C (change lens): Same as base but add "shot on medium format, 80mm equivalent, extremely shallow depth of field, background compressed."

Variant A shifts the mood from clean commercial to high-contrast editorial. Variant B moves the whole shoot from studio to location without touching the character or styling. Variant C changes the optical feel to a luxury fashion aesthetic that reads as aspirational rather than commercial. Same subject, same outfit, three distinct lookbook chapters.

Avoid 🚫

Don’t stack too many stylistic modifiers in a single prompt. “Harsh side lighting, neon color, bokeh, film grain, HDR, hyper-realistic, cinematic” in one prompt creates contradictory signals and Nano Banana 2 tends to average them into something mediocre. Pick two or three strong style anchors and let them do the work.

Build the Lookbook, Then Iterate

The real value of Nano Banana 2 for fashion isn’t any single stunning image — it’s the speed of iteration. A lookbook that would take two days of studio time and a four-person crew to shoot can now be prototyped in an afternoon, presented to a client or creative director, and revised before any real production budget gets committed. The subject consistency feature makes that prototype genuinely useful rather than a rough concept sketch — when your fictional model looks the same across twelve shots, stakeholders can actually respond to the styling and direction rather than mentally compensating for technical inconsistency. Generate the lookbook first, then decide which concepts are worth shooting for real. That’s the workflow that makes sense in 2026.

promptyze

See Full Bio