Features Overview

Go Bananas! provides a comprehensive set of features for AI-powered image generation and management.

Feature Categories

Models & Best Practices

Feature	Description
Pro Model Best Practices	Google's 10 tips for professional generation
OpenAI gpt-image-2	OpenAI's flagship image model — text rendering, prompt adherence, quality tiers
OpenAI GPT Image 2.5	Flare and Sunburst — xhigh/max quality and transparent backgrounds

Golden Rules of Prompting

1. Edit, Don't Re-roll - Use continue_editing for 80%+ correct images
2. Natural Language - Full sentences, not tag soups
3. Materiality - Describe textures: "brushed steel", "soft velvet"
4. Context - Add "for whom": changes everything
5. Identity Locking - "Keep facial features exactly the same as Image 1"

Image Creation & Editing

Feature	Description
Image Generation	Create images from text with style control
Image Editing	Conversational editing without re-uploading

Library Management

Feature	Description
Characters	Consistent characters across scenes
Products	Marketing images with product consistency
Workflow Diagrams	Mermaid → professional visual diagrams & infographics
Style Presets	Reusable style templates

Session & Analytics

Feature	Description
Sessions	Conversation context for workflows
Analytics	Usage tracking and quota management

Feature Comparison

Six models from two providers, one API. Pick per request via `model_id`.

Provider Comparison

Capability	Lite (Nano Banana 2 Lite)	Flash (Nano Banana 2)	Pro (Go Bananas! Pro)	OpenAI gpt-image-2	OpenAI GPT Image 2.5 (Flare, Sunburst)
Default	✅	—	—	—	—
Provider	Gemini 3.1 Flash Lite	Gemini 3.1 Flash	Gemini 3 Pro	OpenAI	OpenAI
Generation speed	Fastest	Fast	Quality	Quality	Flare: fast; Sunburst: precise
Max prompt length	16,384 chars	16,384 chars	16,384 chars	32,000 chars	32,000 chars
Max reference images	14	14	14	16	16
Resolution options	1K	0.5K / 1K / 2K / 4K tiers	1K / 2K / 4K tiers	Any size meeting constraints (≤3,840px edge, multiples of 16, ≤3:1)	Same as GPT Image 2
Aspect ratios	13 standard/cinematic	14 API ratios, including 1:4–8:1 extremes	10 API ratios, including 21:9 and social formats	13 (incl. 21:9 cinematic, 4:5/5:4 social)	13 (same as GPT Image 2)
Output formats	PNG	PNG	PNG	PNG, JPEG, WebP + output_compression	PNG, JPEG, WebP + output_compression
Quality levels	auto	auto	auto	low / medium / high / auto	low / medium / high / xhigh / max / auto
Google Search grounding	No	Yes	Yes	No	No
Thinking mode	Yes	Yes	Yes	Not exposed	Not exposed
Transparent background	No	No	No	Not supported by model	Yes (PNG or WebP)
Text rendering	Basic	Good	Advanced (multilingual)	Best-in-class — small text, dense layouts	Best-in-class
Identity locking	Limited	Strong	Full (up to 5 people)	Strong via reference edits	Strong via reference edits
Default request timeout	30s	60s	120s	240s	240s
Cost	Lowest	Low	Mid	Token-priced by input and output	Token-priced by input and output

Pro Model Exclusive Features

Feature	Description
Text & Infographics	SOTA text rendering for posters, menus, data visualizations
Viral Thumbnails	Identity + text + graphics in one pass
Storyboarding	Multi-scene sequential art with consistent characters
Structural Control	Sketch-to-final, wireframe-to-UI workflows
2D↔3D Translation	Floor plans to renders, 2D art to 3D
High-Fidelity Upscaling	Input images as small as 150x150 enhanced to 4K
Photo Restoration	Repair tears, scratches, fading in old photographs
360 Turnaround	Generate multiple angles from a single reference image
Thinking Mode	Enhanced reasoning for complex prompt interpretation
Google Search Grounding	Real-time data and factual accuracy via web search
Person Generation Safety	Configurable safety modes: dont_allow, allow_adult, allow_all

See Pro Model Best Practices for complete guide.

When to Use Each Model

Use Lite (default) for:

• Quick iterations and drafts
• Simple prompts
• High-volume generation
• Cost-sensitive workflows

Use Flash for:

• Broader aspect ratios
• 2K/4K output
• Stronger multi-reference workflows
• Search-grounded generation

Use Pro for:

• Final production assets
• Print materials (4K resolution)
• Text-heavy content (infographics, posters)
• Multi-person character consistency
• Complex multi-image compositions

Feature Matrix

Feature	MCP Tool	REST API	Web Console
Generate images	✅	✅	✅
Edit images	✅	✅	✅
Continue editing	✅	✅	✅
Characters	✅	✅	✅
Multi-character	✅	✅	✅
Products	✅	✅	✅
Style presets	✅	✅	✅
Search images	✅	✅	✅
Usage stats	✅	✅	✅
Bulk operations	❌	✅	✅
Download ZIP	❌	✅	✅
Admin functions	❌	✅	✅

Quick Examples

Generate an Image

Generate a cozy cabin in snowy mountains at sunset

Edit Conversationally

Add smoke coming from the chimney
Make the sky more dramatic
Add some deer in the foreground

Create and Use a Character

Create a character "Forest Guardian" - an ancient tree spirit
with glowing green eyes and bark-like skin covered in moss

Generate Forest Guardian protecting a hidden waterfall
Generate Forest Guardian teaching woodland creatures

Use a Style Preset

Create style preset "Anime" with prompt prefix:
"anime style, vibrant colors, detailed shading"

Generate a warrior princess using the Anime style

Integration Patterns

MCP Client (Recommended)

Best for conversational AI workflows:

• Claude.ai / Claude Desktop remote connector
• Claude Code
• Codex
• Claude Desktop
• Cursor IDE
• Custom MCP clients

Use remote HTTP OAuth when your client supports it. Use the gobananas-mcp STDIO proxy for command-based or headless clients.

REST API

Best for programmatic access:

• Custom applications
• Batch processing
• Webhooks and integrations

Web Console

Best for visual management:

• Browse and search images
• Manage characters and presets
• View analytics
• Admin operations

Next Steps

Dive into each feature:

1. Pro Model Best Practices → (Start here!)
2. Image Generation →
3. Image Editing →
4. Characters →
5. Products →
6. Workflow Diagrams →
7. Style Presets →
8. Sessions →
9. Analytics →