Example inputhover

prompt: "Design a modern, stylish tech office branded for "each::labs", using the provided each::labs logo and brand colors, with floor-to-ceiling windows overlooking San Francisco and the Golden Gate Bridge. The office must be completely empty — no people, no humans, no staff or visitors anywhere in the scene.Through the large windows, show a stunning panoramic view of the Golden Gate Bridge and the San Francisco bay at golden hour, warm sunset light pouring into the office.Place the full each::labs logo prominently on the main reception wall behind the front desk — it must read the complete wordmark "each::labs" in white, exactly as in the provided logo image, with the signature "::" rendered as four small orange-red squares (two stacked pairs), not circles. Do not crop, cut off, or alter the logo — show the entire "each::labs" name clearly and correctly.Incorporate the brand's color palette — deep navy/midnight blue, crimson red, with the orange-red accent — throughout the space: accent walls, lighting, furniture, and decor. Include a sleek marble reception desk, modern lounge seating, glass meeting rooms, indoor plants, polished concrete floors, and warm modern lighting.The mood is professional, premium, innovative — a high-end AI/tech startup headquarters in San Francisco, shown empty with no people, with the Golden Gate Bridge visible through the windows.Style: photorealistic interior design photography, professional architectural visualization, cinematic golden-hour lighting, high detail, 4K."
num_images: 1
aspect_ratio: "16:9"
output_format: "png"
image_urls
resolution: "1K"
limit_generations: true

Nano Banana 2 Lite · Edit

Array·nano-banana-2-lite·by Google

Nano Banana 2 Lite brings next-generation image editing that's fast, cost-efficient, and delivers sharper, higher-quality edits with rapid generation.

Try it now →

API reference

Runtime (p50): 10s
Estimated price: Usage-based

Call the API

prediction.sh

curl -X POST \
  -H "X-API-Key: $EACHLABS_API_KEY" \
  -H "Content-Type: application/json" \
  --data '{
    "model": "nano-banana-2-lite-edit",
    "version": "0.0.1",
    "input": {
        "prompt": "Design a modern, stylish tech office branded for \"each::labs\", using the provided each::labs logo and brand colors, with floor-to-ceiling windows overlooking San Francisco and the Golden Gate Bridge. The office must be completely empty — no people, no humans, no staff or visitors anywhere in the scene.Through the large windows, show a stunning panoramic view of the Golden Gate Bridge and the San Francisco bay at golden hour, warm sunset light pouring into the office.Place the full each::labs logo prominently on the main reception wall behind the front desk — it must read the complete wordmark \"each::labs\" in white, exactly as in the provided logo image, with the signature \"::\" rendered as four small orange-red squares (two stacked pairs), not circles. Do not crop, cut off, or alter the logo — show the entire \"each::labs\" name clearly and correctly.Incorporate the brand's color palette — deep navy/midnight blue, crimson red, with the orange-red accent — throughout the space: accent walls, lighting, furniture, and decor. Include a sleek marble reception desk, modern lounge seating, glass meeting rooms, indoor plants, polished concrete floors, and warm modern lighting.The mood is professional, premium, innovative — a high-end AI/tech startup headquarters in San Francisco, shown empty with no people, with the Golden Gate Bridge visible through the windows.Style: photorealistic interior design photography, professional architectural visualization, cinematic golden-hour lighting, high detail, 4K.",
        "num_images": 1,
        "aspect_ratio": "16:9",
        "output_format": "png",
        "image_urls": [
            "https://cdn-us.eachlabs.ai/defaults/707c9c9f827b4be8870350c4c604e060.png"
        ],
        "resolution": "1K",
        "limit_generations": true
    },
    "webhook_url": ""
}' \
  https://api.eachlabs.ai/v1/prediction/

Documentation8 sections

Overview
Nano Banana 2 | Lite | Edit Overview

Nano Banana 2 | Lite | Edit is Google’s fast image-editing model for turning an uploaded image into a modified result through natural-language instructions. It is designed for users who need quick visual iteration, not just one-off generation, and it emphasizes speed, cost efficiency, and reliable edit control. Google positions it as Gemini 3.1 Flash-Lite Image, which is the official family name behind this model. Within the Nano Banana family, this model stands out for rapid editing workflows, strong character consistency, and legible text rendering while keeping turnaround times low. It is a practical fit for each::labs users who need the Nano Banana 2 | Lite | Edit API for fast drafts, ad variations, concept edits, and high-volume image-to-image tasks.
Capabilities
Capabilities
- Edits uploaded images using natural-language instructions.
- Generates images quickly, with outputs in as little as four seconds.
- Maintains character consistency across rapid edits and variations.
- Produces legible in-image text for ads, labels, and localized creative.
- Supports contextual scene drafting with real-world knowledge.
- Handles rapid iteration for concept exploration and A/B testing.
- Works as a fast, cost-efficient option in the Google image-to-image family.
- Fits high-throughput developer workflows through the Nano Banana 2 | Lite | Edit API.
Use cases
Use Cases for Nano Banana 2 | Lite | Edit

For creators, this model is useful for turning a selfie or portrait into multiple polished variations while keeping the subject recognizable. A practical prompt is: “Keep the face unchanged, switch the outfit to a denim jacket, and place the subject in a minimal studio background.” Google specifically highlights character consistency and fast iteration for this kind of workflow.

For marketers, it can rapidly produce ad concepts and localized text-heavy visuals. Example: “Edit this banner to feature the same product, add the headline ‘Summer Sale,’ and keep the layout clean for mobile ads.” Its strength in quick text rendering makes it useful for fast creative testing.

For designers, it supports quick mockups and layout exploration. Example: “Convert this rough app screenshot into a polished landing-page hero image with clear hierarchy and modern lighting.”

For developers building on each::labs, it suits high-volume image-to-image pipelines where speed and scale matter. Example: “Use this reference image, preserve the subject, and create three background variants for a product catalog.”
Tips & tricks
Tips and Tricks

For best results with Nano Banana 2 | Lite | Edit, write edit prompts that are specific about what should change and what should stay fixed. Google highlights strong prompt adherence and character consistency, so use instructions that name the subject, the target style, and the exact edit boundary. When editing product shots or creator portraits, mention background, lighting, and framing explicitly. Keep prompts concise when you want fast iterative drafts, then refine with a second pass.

Example prompts:
- “Keep the same person and replace the background with a bright studio set, preserving facial identity and clothing.”
- “Edit this poster to add bold white headline text at the top and keep the brand colors consistent.”
- “Turn this outdoor product photo into a clean ecommerce hero image with centered composition and soft shadow.”
This model works especially well when the source image already contains the right subject and structure, because the edit instruction can be more precise than a full regeneration.
Technical spec
Technical Specifications
- Model family: Gemini 3.1 Flash-Lite Image, branded as Nano Banana 2 Lite.
- Primary mode: Image generation and image editing; this page focus is image-to-image editing.
- Latency: Google says it can generate an image in as little as 4 seconds.
- Cost position: Described by Google as its most cost-efficient image model in the Nano Banana family.
- Inputs: Natural-language prompt plus an uploaded image for edit workflows.
- Outputs: Edited images with prompt-driven changes and preserved subject continuity.
- Formats / aspect ratios: No specific file-format or aspect-ratio limits were confirmed in the search results.
- Architecture details: Google describes it as a Flash-Lite image model, optimized for high-throughput, low-latency use.
Things to be aware of
Things to Be Aware Of

Nano Banana 2 | Lite | Edit is optimized for speed, so it is not the best choice when maximum visual fidelity is the priority. Google notes that editing can take slightly longer than image generation, even though generation itself can arrive in about four seconds. Users should be careful not to over-specify conflicting instructions, because fast models can struggle when too many changes compete in one prompt. If the input image is low quality, heavily cropped, or visually ambiguous, results may be less stable. For best outcomes, start from a clear source image and make one targeted edit at a time.
Key considerations
Key Considerations

Nano Banana 2 | Lite | Edit is best when speed, iteration, and cost matter more than maximum visual richness. Google recommends it for rapid-fire workflows such as prototyping, ad testing, and applications that need to generate many images quickly. It is also a strong choice when you need quick text rendering, character consistency, and contextual edits from a reference image. If your use case depends on highly detailed art direction or slower, higher-end production work, a heavier model may be more appropriate. For most editing tasks on each::labs, the main tradeoff is clear: lower latency and lower cost in exchange for a lighter-weight generation tier.
Limitations
Limitations

Google’s public materials do not confirm fixed output resolutions, file-format restrictions, or aspect-ratio limits for Nano Banana 2 | Lite | Edit in the sources reviewed. The model is designed for rapid editing rather than maximum production-grade detail, so fine texture work and highly complex scenes may be less consistent than on heavier tiers. It also depends on a usable reference image for image-to-image workflows, so it is not a substitute for fully unconstrained text-only generation in every case.