Mastering Nano Banana 2 Image-to-Image Prompts for Product Recoloring

Product photography often requires rapid variations to showcase different colorways without reshooting. Nano Banana 2 enables this through its image-to-image workflow, allowing users to recolor items while preserving the original scene's integrity. This guide explains how to structure your prompts effectively to achieve consistent results when changing product hues.

Understanding the Prompt Structure for Color Changes

When using Nano Banana 2 for recoloring, the prompt must be explicit about what changes and what stays the same. The model interprets instructions based on the relationship between the input image and the text description. To successfully recolor a product, you should avoid vague terms like "change the look" or "make it new." Instead, focus on specific attributes.

A robust prompt structure typically follows this pattern: [Action] + [Target Object] + [New Attribute] + [Preservation Constraints]. For example, rather than saying "make the bottle blue," a more effective instruction would be "recolor the glass bottle to deep navy blue while keeping the liquid level and label position identical." This approach helps the AI understand that the geometry and lighting are fixed, but the surface color is the variable.

It is important to remember that prompt instructions describe desired outcomes; they do not guarantee identity, label, object, or typography preservation. If your product has specific text or logos, the AI may alter them during the recoloring process unless explicitly protected by detailed negative constraints or strong positive reinforcement in the prompt. Always treat the output as an example of the tool's capability rather than a guaranteed final asset.

Step-by-Step Workflow for Consistent Recoloring

To execute a successful recoloring task in Nano Banana 2, follow these structured steps to ensure the best possible alignment with your creative goals.

1. Select the Correct Model: Navigate to the Nano Banana 2 interface at /nanobanana2. Ensure you are using the standard Nano Banana 2 model (Gemini 3.1 Flash Image) rather than the Lite version. Google describes Nano Banana 2 Lite as focused on speed and cost, noting it is not optimized for multiple reference inputs or multi-turn sequential editing. For precise product work requiring high fidelity, the standard model is recommended.
2. Upload Your Base Image: Upload a high-quality product photo where the lighting and shadows are clearly defined. The quality of the input image significantly influences the ability of the AI to maintain texture consistency.
3. Draft Your Prompt: Construct your prompt using the structure discussed above. Be specific about the material. For instance, if recoloring leather, specify "matte black leather" rather than just "black." This helps the model retain the tactile feel of the material.
4. Execute and Review: Submit the request and review the generated images. Check if the lighting direction remains consistent with the original photo. Inconsistent lighting can make the recolored product look pasted onto the background.
5. Iterate if Necessary: If the color bled into the background or the texture was lost, refine your prompt. Add phrases like "maintain original lighting" or "preserve fabric texture" to guide the next generation.

Evaluating Results and Troubleshooting Common Issues

Judging the success of a recoloring prompt involves checking three key areas: color accuracy, texture retention, and lighting consistency. A good result will show the new color applied only to the target object, with no color bleeding onto adjacent elements like hands, tables, or backgrounds. The material properties, such as glossiness or roughness, should remain unchanged.

If the result looks flat or the texture is missing, try adding descriptive words related to the finish, such as "glossy," "matte," "brushed metal," or "woven fabric." If the lighting appears inconsistent, explicitly state "keep original shadow direction" in your prompt.

Sometimes, the AI might struggle to keep the exact shape of the product. This is a known limitation where prompt instructions do not guarantee object preservation. In such cases, consider adjusting the strength of the image influence if the tool allows, or simply accept that the output is an illustrative example of the concept rather than a production-ready file.

For those looking to explore further capabilities, you can Try Nano Banana to experiment with these prompt structures directly. Remember that while the tool offers powerful features, the quality of the output depends heavily on the clarity and specificity of your written instructions.

By following these guidelines and understanding the limitations of the current models, you can streamline your product photography workflow and generate diverse color variations efficiently.