Source photoAI output
Scale Shift
Turn a bookshelf plush into a city-sized movie monster
Keeps the pink fur, round face, paw pads, and the handheld game console, while the shelf becomes a cinematic street scene seen from below.
Google Gemini Flash Image Architecture
Nano Banana 2: Multi-Reference AI Photo Editing & Fusion
Powered by Google's fast multimodal image architecture, Nano Banana 2 is built for combining up to 14 reference photos into a single coherent scene while keeping each person's face, expression, and lighting recognizable.
- Architecture
- Google Gemini Multimodal Image Engine
- Reference Inputs
- Up to 14 reference images (PNG, JPEG, WebP)
- Identity & Pose Lock
- High multi-subject facial consistency & spatial awareness
- Output Resolutions
- 1K, 2K, and 4K with flexible aspect ratios (1:1 to 21:9)
3 Reference Portraits
Nano Banana 2 Image-to-Image
Run Nano Banana 2 in Studio
Upload your photo once, pick Nano Banana 2, GPT Image 2.5, Qwen Image Edit, or Seedream 5.0 in the model selector, and compare outputs directly.
Nano Banana 2 Visual Showcase
What Nano Banana 2 does best
Explore real reference-to-output transformations demonstrating scale shifts, multi-person scene fusion, and 3D social card restyling.
Source photoAI output
Multi-Person Fusion
Combine 3 individual portraits into a lively house party
Preserves each subject's facial structure, hairstyle, and personality while synthesizing unified indoor party lighting.
Source photoAI output
Creative Social Framing
Phone-screen 3D pop-out from a café snapshot
Places the photo on a phone camera screen and lets the subject and her plate break out past the edge, keeping her face, glasses, and top.
Nano Banana 2 Workflow
How to get the best results with Nano Banana 2
Follow these four steps when running single-image edits or multi-reference fusions.
- 01
Upload 1 to 14 reference images
Select Nano Banana 2 in the studio and upload clear front- or three-quarter-angle photos of each subject or object you want to combine.
- 02
Assign roles to each reference
State clearly that the people or products from the uploaded images must keep their exact facial features, logos, or proportions.
- 03
Describe the shared environment and action
Specify where the subjects are located, how they interact, and the unified lighting direction.
- 04
Select aspect ratio and resolution
Start at 1K for fast composition checks, then switch to 2K or 4K once you are happy with the pose and facial likeness.
Reviewed byImg2Img Product Team
Published / Updated /
ReferencesOpenAI Image Generation GuideGoogle Gemini Image Generation Docs
Nano Banana 2 Prompt Recipes
Copy-ready prompts tuned for Nano Banana 2
Click any prompt below to pre-fill the studio workspace.
Multi-Photo Group Shot
Load Nano Banana 2 promptCombine separate portraits into one team or family photo
Combine the people from the uploaded reference photos into a single warm café group portrait sitting at a wooden table, preserving each person's exact facial identity, glasses, and hairstyle.
Works best with 2–4 well-lit portrait uploads.
Product + Model Fusion
Load Nano Banana 2 promptPlace an uploaded product into a lifestyle portrait
Keep the exact person from reference image 1 and the exact perfume bottle from reference image 2. Show the person holding the bottle in soft morning window light with shallow depth of field.
Anchors both the model's identity and the product packaging.
Fast Local Relight
Load Nano Banana 2 promptGolden-hour portrait relighting
Keep the exact face, pose, and outfit from the uploaded photo. Change the lighting to warm golden-hour sunlight with soft rim light on the hair and a blurred sunset city background.
Fast single-reference atmospheric edit.
Explore Workflows & Models
Switch between specialized editing workflows and reference-ready models
OpenAI Multimodal Image Architecture
Explore model guideGPT Image 2.5
Interior redesign, commercial product scenes, complex prompt-guided restyling
Alibaba Qwen Vision-Language Editing Architecture
Explore model guideQwen Image Edit
Targeted hair/outfit edits, vintage photo restoration, sign & packaging text edits
ByteDance Seedream 5.0 Visual Generation Engine
Explore model guideSeedream 5.0
Ink-wash & anime restyling, high-fashion editorial portraits, 4K visual campaigns
Black Forest Labs Flux Architecture
Explore model guideFlux Kontext
Commercial advertising, brand packaging preservation, architectural concept design
xAI Grok Visual Intelligence Engine
Explore model guideGrok Imagine AI
Viral social media cards, 3D pop-out framing, bold retro pixel-art transformations
Prompt-Guided AI Image Editor
Open in StudioAI Image Editor
Edit photos with natural language prompts using our free AI image editor. Restore old photos, change outfits, relight portraits, and swap product scenes while keeping the original composition.
AI Photo to Sketch Converter
Open in StudioPhoto to Sketch AI
Convert any photo to pencil sketch, charcoal drawing, ink-wash painting, or manga line art online with AI, keeping the likeness, pose, and composition recognizable.
AI Old Photo Restoration & Colorization
Open in StudioOld Photo Restoration
Restore damaged, scratched, and faded vintage photos online with AI. Remove paper creases, glare, and tears while colorizing monochrome portraits with natural skin tones.
AI Shadow Remover & Portrait Relighting
Open in StudioRemove Shadow from Photo
Remove unwanted facial shadows, background glare, and hard lighting shadows from photos online with AI. Relight portraits and product shots while keeping the original subject recognizable.
AI Filter Remover & Color Neutralization
Open in StudioRemove Filter from Photo
Remove Instagram, TikTok, and oversaturated color filters from photos online with AI. Neutralize warm orange and magenta color casts to restore authentic skin tones and clean lighting.
AI Watermark Remover & Inpainting
Open in StudioAI Watermark Remover
Remove watermarks, logos, timestamps, and unwanted text from photos online with AI. Seamlessly reconstruct hidden background textures without blur.
AI Image Extender & Outpainting Tool
Open in StudioAI Expand Image
Expand photos and uncrop images beyond their original borders online with AI. Extend vertical 9:16 portraits to 16:9 widescreen wallpapers with seamless continuity.
Nano Banana 2 FAQ
Frequently asked questions about Nano Banana 2
Key details on reference image limits, resolution options, and ideal workflows.
What is Nano Banana 2 used for in image-to-image generation?
Nano Banana 2 is a fast multimodal image-to-image model designed for editing existing photos and combining up to 14 reference images—such as merging separate portraits into a group scene or pairing a product with a background reference.
When should I choose Nano Banana 2 over other models?
Pick Nano Banana 2 when one result has to combine several reference photos, such as people from separate shots in one group scene. For a single local change like hair color or sign text, Qwen Image Edit is usually more targeted; for a full style change, try Seedream 5.0.
How many credits does Nano Banana 2 use?
Nano Banana 2 starts at 11 credits per image. The studio shows the estimate for your settings before you generate, and failed generations return the credits.
How many reference images can I upload to Nano Banana 2 at once?
You can upload between 1 and 14 reference images (up to 10 MB per image) in PNG, JPEG, or WebP format.
What resolutions and aspect ratios does Nano Banana 2 support?
Nano Banana 2 supports 1K, 2K, and 4K resolutions and flexible aspect ratios ranging from 1:1, 4:3, and 16:9 to ultra-wide 21:9.
How does Nano Banana 2 compare to Qwen Image Edit?
Nano Banana 2 is stronger when merging multiple reference images into a new scene, whereas Qwen Image Edit specializes in surgical single-image modifications and text replacement.