---
title: Nano Banana Mastery
category: product
entity_type: skill
price: $15
canonical: https://forgehouse.ai/skills/nano-banana-mastery/
lang: en
hreflang_alt: https://forgehouse.ai/tr/skiller/nano-banana-mastery/
last_updated: 2026-06-20
---

# Nano Banana Mastery

> Master Gemini 3 Pro Image (AI tools) prompt engineering for production-grade social media…

Production-grade prompt engineering for Gemini 3 Pro Image (Nano Banana) so your social media creative comes out sharp and brand-consistent instead of plastic and AI-looking. It teaches the six-part prompt anatomy: subject, composition, style, lighting, mood, negative: plus aspect-ratio locking, negative baselines, text-via-quoting, character consistency, and a draft-refine-polish chain that cuts iteration cost while raising quality.

## Use cases
- Generating Instagram feed, Story, and Reels post visuals
- Producing service announcements and testimonial cards
- Keeping one character consistent across a series of posts
- Rendering readable text overlays via the quoting technique
- Anchoring brand style from a moodboard reference image
- Building event posters and product hero images per platform

## Benefits
- Cut iteration spend with a fast-then-quality refinement chain
- Eliminate common artifacts like extra fingers with a negative baseline
- Keep a recurring character recognizable across an entire series
- Lock every output to the correct aspect ratio for its platform

## What’s included
- Six-component prompt anatomy with a full worked example
- Mandatory negative baseline and aspect-ratio lock pattern
- Three-stage draft, refine, and polish chain with quality modes
- Character consistency and moodboard style-blending patterns
- Quote-format text integration for legible overlays
- Sector-specific negative injection for regulated industries

## Who it’s for
For agencies, social media managers, and creators who need consistent, brand-compliant, platform-ready image creative without endless costly retries.

## How it runs
Roughly 43 percent: that is what a three-stage prompt chain saves versus retrying in quality mode, and that economy is the backbone of this production loop for social media creative.
1. Decomposes every prompt into 6 components (SUBJECT, COMPOSITION, STYLE, LIGHTING, MOOD, NEGATIVE) at a 60-80 word optimum, so each part can be tested and fixed in isolation instead of throwing the whole prompt away.
2. Locks the aspect ratio explicitly per platform (1:1 feed, 4:5 carousel, 9:16 story) and appends a 15-word negative baseline to every call, killing the classic artifacts: extra fingers, plastic skin, watermarks, gibberish text.
3. Runs chain prompting instead of one-shot generation: a fast-mode draft for composition check, a balanced-mode refine for style and detail, a quality-mode polish for the final. The 3-stage chain cuts iteration cost roughly 43 percent versus retrying in quality mode.
4. Renders overlay text with strict quoting discipline: the exact text in double quotes plus font family, weight and color. Turkish diacritic text is routed to the HTML-to-PNG pipeline instead, where lettering is guaranteed.
5. Keeps series posts consistent by enabling maintainCharacterConsistency and holding the character description identical across generations; only the action and scene change between posts.
6. Injects sector-specific legal negatives for regulated clients (no medical claims, no weight numbers, no income promises, no before-after comparisons) and logs successful prompts back into the pattern library with cost and quality metadata.

## FAQ
### Do these prompt patterns work in Midjourney or DALL-E too?
The six-part anatomy thinking transfers loosely, but the specific patterns are tuned to Gemini 3 Pro Image: the aspect-ratio lock, the quote-format text technique, and the draft-refine-polish quality modes all rely on how this model behaves. Expect to re-derive them elsewhere, not copy-paste.

### How does it get legible text into an image when AI text is usually garbled?
Through the quoting technique: the exact overlay string is wrapped in quotes inside the prompt, which this model treats as text to render rather than a vague concept to illustrate. Combined with the negative baseline, it cuts the classic mangled-letters failure substantially.

### Will every output come out on-brand without review?
No. Brand-parameter injection and moodboard anchoring narrow the drift, and the refinement chain reduces wasted retries, but image models still produce variance. Plan on a human pass to select and reject; the skill lowers the cull rate, it doesn't remove the cull.

## Price
$15, one-time, no subscription. VAT included.

Related guide: [Building a design system with Claude Design](https://forgehouse.ai/guides/claude-design-design-system/)