Mastering Underwater Refraction: Nano Banana 2 Image-to-Image Prompt Guide

Understanding the Goal of Underwater Simulation

Transforming a standard photograph into an underwater scene requires more than simply adding a blue tint. The core challenge lies in replicating the optical physics of water, specifically how light bends and scatters as it passes through a denser medium. When using Nano Banana 2 for this environment change, the objective is to introduce visual cues that convince the viewer the subject is submerged. These cues include the distortion of straight lines, the appearance of shimmering light patterns known as caustics on surfaces, and a shift toward cooler, desaturated colors due to the absorption of red wavelengths.

It is important to note that prompt instructions describe desired outcomes but do not guarantee identity, label, object, or typography preservation. While the tool aims to maintain the essence of the original image, the intensity of the refractive effects may alter specific details. Users should approach these edits with the understanding that the AI interprets descriptive language to reconstruct the scene rather than applying a static filter. This distinction is crucial when setting expectations for complex environmental changes like submersion.

Structuring Your Prompt for Optical Effects

To achieve realistic results, the prompt must be structured to prioritize physical phenomena over artistic style. A generic request for "underwater" often yields a flat blue overlay. Instead, the prompt needs to explicitly describe the interaction between light and the subject. Begin by defining the medium, then layer in specific optical behaviors. For instance, instead of saying "make it look like water," specify "simulate light bending through water" or "add caustic light patterns."

Consider the following example structure for your input. This format helps the model understand the hierarchy of effects:

1. Base Context: State the current state of the image (e.g., "a person standing on a beach").
2. Environmental Shift: Define the new medium (e.g., "submerged in clear ocean water").
3. Optical Physics: Detail the refraction and lighting (e.g., "distorted limbs due to refraction," "shimmering caustic shadows on skin," "loss of red tones, shift to cyan and teal").
4. Atmosphere: Add depth cues (e.g., "slight particulate haze," "soft focus from water density").

Remember that these are examples of prompt structures. You can adapt them based on your specific source image. The goal is to provide enough descriptive data for the model to calculate the necessary pixel adjustments without overwhelming the system with contradictory instructions. If you are using Nano Banana 2 Lite, keep in mind that it is focused on speed and cost and is not optimized for multiple reference inputs or multi-turn sequential editing. Complex refraction tasks may yield better consistency on the standard Nano Banana 2 workflow.

Step-by-Step Workflow for Image-to-Image Editing

Executing this transformation involves a systematic approach within the Nano Banana 2 interface. Follow these numbered steps to ensure you utilize the image-to-image capabilities correctly:

1. Upload Source Image: Navigate to the Nano Banana 2 product page at /nanobanana2 and select the image-to-image mode. Upload the photo you wish to transform.
2. Input the Prompt: Enter your structured prompt describing the underwater refraction effects. Ensure you clearly articulate the need for light bending and color shifting.
3. Adjust Strength Parameters: If available, modify the influence strength. A higher strength will apply the water effects more aggressively, potentially distorting the original geometry more significantly. Lower strength preserves more of the original detail while hinting at the environment.
4. Generate and Review: Run the generation process. Observe how the model handles the edges of the subject and the background. Check if the caustics appear natural or if the distortion looks artificial.
5. Iterate if Necessary: If the result lacks realism, refine the prompt. Add specific terms like "wavy surface distortion" or "light scattering" to guide the next iteration.

For users seeking a quick start, Try Nano Banana offers a direct path to begin experimenting with these prompts immediately.

Judging Results and Troubleshooting Common Issues

Evaluating the success of your underwater simulation requires looking for specific visual markers. A successful edit will show smooth transitions where light hits the subject, mimicking the way water magnifies and warps objects. The colors should naturally fade from warm tones near the top to cooler blues deeper down, assuming a vertical composition. If the image appears too flat or the subject looks like they are wearing a blue mask, the prompt likely lacked sufficient detail regarding refraction and light scattering.

Common issues include excessive blurring or loss of facial features. This often happens when the prompt emphasizes "haze" too heavily without balancing it with sharpness descriptors. To fix this, rephrase the prompt to focus on "clear water" or "crisp refraction" rather than just "murky depths." Another frequent problem is the failure to generate caustics. If no light patterns appear, explicitly add "dynamic caustic light patterns on the subject" to the instruction set.

Finally, remember that Google documents Nano Banana 2 as Gemini 3.1 Flash Image, and different models have distinct capabilities. Do not assume features available in one version exist in another without verification. By focusing on the physics of light and water in your prompt construction, you can guide the AI to produce convincing underwater scenes that feel authentic and immersive.