Nano Banana 2 Reference Image Checklist for E-Commerce Consistency

Creating a cohesive e-commerce catalog requires more than just generating individual images; it demands strict consistency across an entire batch. When using Nano Banana 2 (identified by Google as Gemini 3.1 Flash Image), the quality of your output relies heavily on how you prepare your input materials. This tool excels at image-to-image workflows, but without a structured approach, the AI may introduce unwanted style shifts, altering backgrounds or lighting between items.

This guide provides a step-by-step checklist to curate reference images that maintain uniformity. By following these protocols, you can ensure that every product in your batch looks like it belongs to the same collection, shot under identical conditions.

The Core Prerequisites for Consistent Outputs

Before initiating any generation task, you must establish a baseline for what "consistent" means for your specific brand. The AI does not inherently know your visual standards unless explicitly guided through your reference inputs. The primary goal is to minimize variables that the model might interpret as creative license.

First, ensure you are selecting the correct model tier. While Nano Banana 2 Lite focuses on speed and cost efficiency, it is not optimized for multiple reference inputs or complex multi-turn sequential editing. For tasks requiring high-fidelity consistency across a batch, relying on the standard Nano Banana 2 capabilities is advisable. Do not assume features available on other platforms apply here without verification.

Second, gather your raw assets. These should be generic, unbranded product shots if you are testing styles, or actual product photos if you are refining existing catalogs. Remember that Nano Banana refers to the AI image generation tool, not a physical cosmetic brand or bottle. Your prompts will describe desired outcomes, but they do not guarantee identity preservation of specific labels or typography. Therefore, the visual consistency must come from the reference images themselves rather than text instructions alone.

A Step-by-Step Workflow for Curating References

To achieve professional results, follow this numbered workflow to prepare your dataset before entering the generator.

1. Standardize the Background: Select reference images where the background is neutral and uniform. If your products require a white backdrop, ensure all source images have clean, consistent white tones without shadows or gradients. Avoid mixing studio shots with lifestyle backgrounds in the same batch.
2. Align Lighting Direction: Check the direction of light sources in your reference set. Shadows should fall consistently (e.g., always from the top-left). If one image has harsh overhead lighting and another has soft side lighting, the AI may struggle to blend them into a single aesthetic.
3. Fix Perspective and Angle: Ensure all reference images share the same camera angle. A front-facing view should not be mixed with a 45-degree angle shot unless you intend to create variety. Consistency in perspective is crucial for catalog grids.
4. Verify Resolution and Aspect Ratio: Resize all reference images to the same dimensions. Mismatched aspect ratios can cause the AI to crop or stretch elements unpredictably during the generation process.
5. Select the Primary Anchor: Choose one high-quality image to serve as the primary anchor for the batch. This image sets the tone for color grading, texture, and overall mood. Use this as the main input for subsequent generations.

Once your references are curated, you can proceed to generate. You can copy example prompts from the prompt library to get started, treating them as templates rather than guaranteed solutions. For instance, you might use a prompt describing a "clean white background with soft studio lighting," but remember that these instructions describe desired outcomes and do not guarantee specific object preservation.

Try Nano Banana

How to Judge Results and Fix Common Issues

After generating your batch, evaluate the outputs against your original checklist. Look for subtle shifts in color temperature or background noise. If the AI introduces unwanted style shifts, such as changing a matte finish to glossy, it often indicates that the reference images contained conflicting metadata or visual cues.

If you notice inconsistencies, try the following fixes:

• Refine the Reference Set: Replace any outlier images that deviate from the lighting or angle standard. Sometimes a single inconsistent photo can skew the entire batch.
• Adjust Prompt Specificity: Add more descriptive terms regarding lighting and background to your prompt. However, avoid over-relying on text for details that should be handled by the image input.
• Re-run with Different Models: If the standard model struggles with a specific constraint, consider if a different model tier is appropriate, keeping in mind the limitations of Lite versions regarding multi-turn editing.

It is important to note that while Google documents Nano Banana 2 as Gemini 3.1 Flash Image, the website's specific feature availability may vary. Always verify current capabilities on the product page. Do not treat prompt examples as tested guarantees; they are illustrative tools to help you start. By rigorously preparing your reference images, you empower the AI to deliver the uniform, professional look required for modern e-commerce catalogs.

For further details on image generation documentation, refer to the official resources provided by Google. This ensures you stay updated on the latest model behaviors and capabilities without relying on unverified third-party claims.