ImageComposite

Nano Banana 2 – Composite

Composite model by Google — US.

5/5 LORY rating
About this model

Nano Banana 2 is best for fast, higher-throughput composites and reference merges.

Nano Banana 2 uses Google's Gemini 3.1 Flash Image Preview model. It keeps the fast Banana workflow while expanding reference-image flexibility, making it a good middle ground for character, object, and style merges that need more control than the original Nano Banana.

CapabilityComposite
ProviderGoogle
OriginUS
OutputImage
ModesComposite
Model details

Practical specs for planning a generation in LORY. These details come from the model contract we use when routing a request.

InputsPrompt required · Up to 14 source images · Up to 14 reference images
OutputPNG, JPEG
Resolution512px, 1K, 2K, 4K
Aspect ratios1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1 +7 more

Try Nano Banana 2 – Composite on LORY

Start with free welcome credits — no subscription, and we never ask for payment info during your trial. Pay only when you decide to top up.

Similar modelsMore image models you might like
ImageComposite

Nano Banana – Composite

Nano Banana is a fast choice for simple composites and quick reference-driven blends.

Google · USView details
ImageComposite

Nano Banana Pro – Composite

Nano Banana Pro is a strong choice for high-fidelity composites that keep characters and style more consistent across references.

Google · USView details
ImageEdit

Nano Banana 2 – Edit

Nano Banana 2 is a strong choice for faster edits with better image quality and broader input support.

Google · USView details
ImageText-to-Image

Nano Banana 2 – Generate

Nano Banana 2 balances speed, image quality, and higher-volume creative work.

Google · USView details
ImageComposite

Seedream 5.0 Lite – Composite

ByteDance Seedream 5.0 Lite multi-reference composition — blends up to 10 references with prompt-led restyling.

ByteDance · ChinaView details
ImageComposite

GPT Image 2 – Composite

Blend multiple reference images into a single cohesive scene with GPT Image 2. Inputs are processed at high fidelity.

OpenAI · USView details
Try on LORY