Nano Banana 2 – User guide

Written By Stanislas

Last updated 9 days ago

Overview

Nano Banana 2 is a powerful image generation and editing bot powered by Google's Gemini 3.1 Flash Image model. It transforms text descriptions into high-quality images, edits existing images through natural language prompts, and can access real-time information from the web to ground your creations in current events, weather, and trending topics.

This bot is ideal when you need fast, versatile image generation—whether you're creating product mockups, illustrations, marketing assets, or experimenting with visual ideas. With support for multi-image fusion, character consistency, and advanced grounding options, Nano Banana 2 handles everything from simple text-to-image requests to complex, multi-turn editing conversations.

Prerequisites

Before using Nano Banana 2, ensure you have:

  • Active Swiftask workspace with appropriate plan
  • Access to Chat or Agents (depending on your use case)
  • Workspace credits sufficient for image generation costs (see Pricing & Limits below)

Step-by-Step Guide

Basic: Generate an image from text

  1. Open Chat in your Swiftask workspace.
  2. Select Nano Banana 2 from the AI Tools or Bot selector menu.
  3. Enter your image prompt in the message box. Be descriptive: include style, mood, composition, and any specific details.
  • Example: "A serene mountain landscape at sunset, oil painting style, golden hour lighting, photorealistic"
  1. Press Send or Generate.
  2. The bot will process your request and return a generated image.

Advanced: Add reference images for style transfer

  1. In Chat, select Nano Banana 2.
  2. Click Add Images or Upload References (or attach images to your prompt).
  3. You can add up to 14 reference images to guide style, composition, or character appearance.
  4. Write your prompt, referencing the style or elements you want: "Apply the art style from image 1 to a modern office building"
  5. Select Aspect Ratio (optional):
  • Default: matches input image dimensions
  • Or choose: 1:1, 2:3, 3:2, 4:3, 16:9, 21:9, 1:4, 4:1, 1:8, 8:1, etc.
  1. Select Resolution (optional):
  • 1K (default) – fastest, lowest cost
  • 2K – balanced quality and cost
  • 4K – highest quality, highest cost
  1. Press Generate.

Advanced: Enable real-time web grounding

To make your images reflect current events, weather, or trending topics:

  1. In Chat, select Nano Banana 2.
  2. In the Advanced Options or Settings panel, toggle Google Search Grounding ON.
  3. Write a prompt that references current information: "Generate an image of the top trending fashion at Paris Fashion Week 2025"
  4. Press Generate.
  5. The bot will fetch real-time web data and incorporate it into your image generation.

Note: The first 5,000 requests per month are free. Additional requests incur a charge of $14 per 1,000 requests.

Advanced: Image editing via text prompts

  1. In Chat, select Nano Banana 2.
  2. Attach or upload an existing image.
  3. Describe what you want to change: "Change the background from blue to a sunset sky" or "Make the person's shirt red instead of blue"
  4. Press Generate.
  5. The bot will edit and return the modified image.

For complex edits, you can continue the conversation—describe additional changes and the bot will refine the image iteratively.

Advanced: Use image search grounding

To use web images as visual context for your generation:

  1. In Chat, select Nano Banana 2.
  2. In Advanced Options, toggle Image Search Grounding ON.
  3. Write a prompt that benefits from visual examples: "Create a product mockup inspired by minimalist design trends from Pinterest"
  4. Press Generate.
  5. The bot will search for relevant images online and use them to inform the generation.

Advanced Features

Character consistency

Nano Banana 2 can maintain consistent characters across multiple generations. To use this:

  1. Generate an initial image with your character: "A female warrior with red hair, blue armor, standing pose"
  2. In follow-up prompts, reference the character: "Same warrior, now sitting on a throne" or "Character in a winter landscape"
  3. The bot will keep the character's appearance consistent across generations.

Multi-image fusion (up to 14 references)

You can blend styles, compositions, or visual elements from multiple reference images:

  1. Upload 2–14 reference images.
  2. Write a prompt that combines elements: "Merge the color palette of image 1, the composition of image 2, and the subject from image 3"
  3. Press Generate.

This is powerful for:

  • Creating mood boards
  • Blending artistic styles
  • Compositing product mockups
  • Developing consistent visual themes

Multilingual text rendering

Nano Banana 2 can render crisp, readable text in images in multiple languages:

  1. Include text in your prompt: "Create a poster with the text 'Hello World' in English and '你好世界' in Chinese, centered, bold sans-serif font"
  2. Press Generate.
  3. The bot will render the text cleanly in the image.

Pricing & Limits

Per-image costs

ResolutionCost per image
1K (1024px)$0.067
2K (2048px)$0.101
4K (4096px)$0.151

Token-based pricing

If you're tracking token consumption:

  • Input tokens: $0.25 / 1M tokens
  • Output text tokens: $1.50 / 1M tokens
  • Output image tokens: $60.00 / 1M tokens
  • Swiftask token coefficient: 0.875 (affects how your workspace credits are consumed)

Google Search grounding costs

  • First 5,000 requests per month: Free
  • Additional requests: $14 per 1,000 requests

Limits and quotas

  • Maximum reference images: 14
  • Supported output formats: JPG (default), PNG
  • Popularity: 4+ million runs (highly stable and battle-tested)

Troubleshooting

Generation takes too long

Cause: High demand or complex prompts.

Fix:

  • Reduce resolution to 1K for faster results.
  • Simplify your prompt; very long or complex descriptions may increase processing time.
  • Try again during off-peak hours.

"Google Search grounding limit exceeded"

Cause: You've used more than 5,000 free Google Search requests this month.

Fix:

  • Disable Google Search grounding to continue generating (images will use only your prompt text).
  • Wait until next month for the free quota to reset.

Image doesn't match the prompt

Cause: Prompt was too vague or conflicting.

Fix:

  • Be more specific: include style, mood, lighting, composition, and subject details.
  • Example: Instead of "a cat", try "a tabby cat sitting on a sunny windowsill, photorealistic, warm lighting, cozy home interior"
  • Add reference images to guide the style.
  • Try multiple times—image generation has inherent variation.

Reference images aren't being used

Cause: Prompt doesn't clearly reference the uploaded images.

Fix:

  • Update your prompt to explicitly mention the reference images: "Use the color palette from image 1 and the composition of image 2"
  • Ensure reference images are clear and relevant to your request.

Output format is JPG but I need PNG

Cause: Default output format is JPG.

Fix:

  • In Advanced Options, change Output Format to PNG.
  • Press Generate again.

FAQ

Q: Can I edit generated images?
A: Yes. Upload the generated image and describe the changes you want in your next prompt. You can iterate multiple times.

Q: What's the difference between 1K, 2K, and 4K resolution?
A: 1K (1024px) is fastest and cheapest, ideal for web and social media. 2K (2048px) offers better detail for print or high-quality displays. 4K (4096px) is for maximum detail but is slower and more expensive.