Nano Banana 2 Lite Edit API

google/nano-banana-2-lite/edit
10 ratios · 5 credits ($0.025) / generation

Nano Banana 2 Lite Edit refines images using natural language instructions and 1 to 14 reference images in approximately 4 seconds, supporting multi-reference feature fusion, fast single-step local edits, and flexible aspect ratio control. It preserves original poses, subject identity, and illumination consistency during background swaps or element tweaks while delivering efficient character continuity at just 5 credits ($0.025) per generation.

Get API Key
Input
790
2/14
Reference image 1: Reference 1
Reference image 2: Reference 2

Examples

output.jpg

Use image 1 as the identity and bicycle reference, and image 2 as the destination landscape. Create a seamless documentary photograph of this same adult repairman pushing this same old dark-green bicycle along the broad white salt embankment in image 2. Preserve his face, straw hat, indigo shirt, tan trousers, shoes and the bicycle frame, handlebars and two wheels. Preserve the pink ponds, white salt texture, horizon and perspective of image 2. Place the complete man and bicycle naturally on the foreground embankment, with feet and both tires contacting the ground. Match scale and upper-left sunlight, with physically consistent contact shadows. One man, one bicycle, no collage borders, no duplicate limbs or wheels. No advertisements, brands, logos, signatures, text or watermarks.

output.jpg

Transform only the stage scenery in this theatre photograph into an underwater-inspired theatrical installation. Keep the exact camera view, rectangular proscenium, stage floor boundaries and foreground seating rows. Hang five translucent jellyfish-shaped fabric lanterns at varied heights with visible fine suspension wires. Add layered blue-green gauze backdrops and rippling water-light projections on the existing dry wooden stage. The lanterns are crafted stage props, not living animals. Preserve the theatre structure and empty audience seats, no performers. It must clearly remain a real indoor theatre with a magical practical-lighting set, not become an actual ocean or flooded room. No advertisements, brands, logos, signatures, text or watermarks.

output.jpg

Convert this red-eyed tree frog photograph into a bold four-color screen-printed comic illustration. Preserve the same single frog, exact resting pose, visible toes gripping the banana leaf edge, bright red eyes, green body, flank markings, leaf outline, diagonal composition and original framing. Use confident thick black contour lines, flat cyan-magenta-yellow-black ink separations, visible halftone dots and subtle paper grain. Simplify the background into dark graphic foliage while retaining separation around the frog. This is a single full-frame illustration, with no panels, speech balloons, lettering or added characters. No advertisements, brands, logos, signatures, text or watermarks.

Nano Banana 2 Lite Edit Overview

Nano Banana 2 Lite Edit is a high-speed, lightweight image editing and multi-reference synthesis model engineered by Google. Capable of ingesting up to 14 reference pictures, it executes prompt-directed modifications such as garment updates, background replacements, color shifts, and multi-asset feature blends in roughly 4 seconds. By tightly preserving original poses, facial identities, and ambient lighting harmony, it delivers an agile, cost-effective solution for e-commerce apparel visualization, contextual product relocation, and episodic character development.

Why Choose Nano Banana 2 Lite Edit?

  • Up to 14 Reference Images SupportIngests 1 to 14 reference images in a single call, intelligently decoupling identities, garments, and product textures across multiple visual sources.

  • Fast Single-Step Local Edits in ~4sCompletes background replacements, color adjustments, and localized object swaps in roughly 4 seconds without manual masking or complex layer setups.

  • Faithful Pose, Composition, and Lighting RetentionModifies requested targets while strictly preserving facial features, bodily posture, perspective lines, and ambient contact shadows.

  • Zero Surcharge on Reference ImagesTransparent, predictable pricing of 5 credits ($0.025) per generation regardless of whether you upload 1 or 14 reference images.

  • Robust Character ContinuityMaintains dependable visual identity across successive short-session editing iterations and variations.

Parameters

ParameterRequirementDescription
promptRequired

Non-blank string, at least 1 character after trimming. No upstream maximum length is specified.

image_urlsRequired

1–14 publicly accessible HTTP(S) image URLs.

sizeOptional

Output aspect ratio. Select auto to let the model choose the framing; the API request omits size.

Defaultauto1:12:33:23:44:34:55:49:1616:921:9

How to Use

  1. Supply reference image URLsProvide 1 to 14 publicly accessible HTTP(S) image URLs serving as your base canvas, character reference, or target wardrobe props.

  2. Draft concise editing instructionsState what to alter and what to preserve using explicit syntax such as "Change only the background to an alpine lake, keeping the person and lighting unchanged."

  3. Choose output framingSelect from 10 supported aspect ratios (such as 1:1, 9:16, or 16:9), or select auto to allow adaptive framing.

  4. Dispatch asynchronous edit jobSubmit your API payload to immediately receive a unique task_id while background workers process feature decoupling and diffusion.

  5. Retrieve verified outputPoll the status endpoint until complete, then fetch and download your edited 1K resolution image.

Pricing

Text to Image and Image Edit both cost 5 credits ($0.025) per generation. 1 credit = $0.005; no reference-image surcharge.

UsageRateDetails
Image Edit5 credits / generation$0.025 / generation; Google comparison: $0.034 / generation, about 26% less.

Best Use Cases

  • E-commerce apparel swaps & styling lookbooksAnchor a model's pose and facial identity while swapping tops, dresses, or outerwear using separate clothing reference images.

  • Product re-contextualization & background swapsMove studio white-background packshots into realistic modern living rooms, sunny patios, or urban streets.

  • Rapid product color & material explorationShift a device casing from matte black to silver brushed aluminum to generate quick catalog variations.

  • Storyboard scene variants & asset stagingSwap props or alter background environments behind key characters to build episodic story sequences quickly.

Pro Tips

  • Use explicit boundary syntax: Formulate instructions with clean preservation boundaries, such as "Change only the jacket to a navy wool pea coat, keeping the person's pose, facial expression, and warm studio lighting unchanged."
  • Use unobstructed reference imagery: Provide images with clear lighting and uncluttered backgrounds to facilitate precise subject and texture decoupling by the multimodal encoder.
  • Explicitly reference source images: When providing multiple references, specify their roles clearly (e.g., "Retain the model from image 1 wearing the sunglasses from image 2").
  • Iterate in fast single steps: For complex transformations, perform sequential single-step adjustments to capitalize on the ~4-second turnaround and isolate creative variables.

Notes

  • The image_urls parameter is required and must contain an array of 1 to 14 valid public HTTP(S) image URLs.
  • The prompt parameter is required, non-empty, and must contain at least 1 character after trimming leading and trailing whitespace.
  • Requests operate asynchronously and return a unique task_id for polling execution status and retrieving the finished asset.

Related Models

Nano Banana 2 Lite Edit API frequently asked questions

What is the Nano Banana 2 Lite Edit API?

Nano Banana 2 Lite Edit is a speed-optimized image editing model developed by Google. It pairs 1 to 14 reference images with natural language instructions to execute fast local inpainting, background replacements, and multi-asset feature blends. Powered by Google's streamlined multimodal architecture, it delivers completed edits in approximately 4 seconds while preserving subject identity, bodily posture, and lighting consistency with zero per-image reference surcharges. You can call it programmatically or try it from the playground above.

How many reference images does Nano Banana 2 Lite Edit support?

The model accepts an array of 1 to 14 publicly accessible HTTP(S) image URLs. You can provide base portraits, product photos, and background references simultaneously, allowing the model to intelligently decouple and blend attributes into a unified render.

How does Nano Banana 2 Lite Edit preserve subjects during edits?

The architecture maintains tight spatial anchors for unmentioned regions. Using explicit prompts such as "Change only [X], keeping [Y] unchanged" instructs the model to localize diffusion strictly to the target element while leaving identity, camera perspective, and ambient lighting untouched.

Is there an extra fee for uploading multiple reference images?

No. Unlike models that apply surcharges per input image, Nano Banana 2 Lite Edit charges a flat rate of 5 credits ($0.025) per generation whether you provide 1 reference picture or the full limit of 14 images.

What common editing operations are supported?

The model excels at rapid single-step edits including wardrobe changes, background replacement, color and texture shifts, removing small distracting elements, and inserting new context-appropriate accessories.

How do I configure the output aspect ratio in Edit?

You can specify any of the 10 supported aspect ratios (1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, or 21:9) via the optional size parameter. If omitted, the model generates framing adapted to the source inputs.

What is the recommended prompt formula for Nano Banana 2 Lite Edit?

Follow a structured pattern: identify the target subject, describe the exact transformation, cite any specific reference images, and state preservation constraints (e.g., "Change the subject's shirt in image 1 to the leather jacket in image 2, keeping the pose, facial expression, and studio lighting identical").