Nano Banana 2 Prompt for Removing Text Overlays from Existing Photos

When working with digital photography, unwanted text overlays, watermarks, or captions can sometimes ruin an otherwise perfect image. Whether you are cleaning up a screenshot for a presentation or removing a distracting logo from a stock photo, the goal is to make the alteration look natural. Nano Banana 2, identified as Gemini 3.1 Flash Image, offers powerful capabilities for this task through its image-to-image workflows. By using precise instruction phrasing, users can guide the model to intelligently reconstruct the underlying scene where the text once existed.

It is important to understand that Nano Banana refers to the AI image generation and editing tool itself, not a skincare brand or physical product. The prompts provided here describe desired outcomes but do not guarantee the preservation of specific identities, labels, or exact typography unless explicitly requested. When using these tools, the focus remains on seamless visual restoration rather than copying existing text.

Understanding the Reconstruction Workflow

To effectively remove text, the process relies on the model's ability to infer context. When you upload an image containing text, the system analyzes the surrounding pixels—such as sky textures, fabric patterns, or architectural lines—to determine what should fill the gap left by the removed characters. This is an image-to-image workflow where the original image serves as a reference, and the prompt acts as the directive for modification.

The prompt instructions must be clear about the removal action while describing the background elements that need to be regenerated. For instance, if a watermark sits over a grassy field, the prompt should emphasize the texture of the grass. If the text covers a building facade, the instruction should focus on the brickwork or window details. This approach ensures that the generated content blends naturally with the rest of the photograph.

Users should note that while the prompt library offers example prompts that can be copied, these serve as starting points. Adjustments may be necessary depending on the complexity of the image. Furthermore, Google documents Nano Banana 2 Lite as focused on speed and cost, noting it is not optimized for multiple reference inputs or multi-turn sequential editing. Therefore, for complex text removal tasks requiring high fidelity, the standard Nano Banana 2 model is generally more suitable than the Lite version.

Five Materially Different Prompts for Text Removal

Below are five distinct prompt examples designed for different scenarios. These are labeled as examples to illustrate how phrasing changes based on the subject matter. Each prompt targets a specific type of overlay and background reconstruction.

1. The Minimalist Landscape Cleanup

Use Case: Removing a small date stamp or camera metadata from a nature photograph where the background is uniform, such as a clear sky or calm water. Prompt Example: "Remove the white date stamp in the bottom right corner. Reconstruct the smooth blue sky texture behind it seamlessly." Adjustment: If the sky has clouds, add "match the cloud density and direction" to the prompt to ensure the reconstructed area looks consistent with the surrounding atmosphere.

2. The Urban Architecture Restoration

Use Case: Erasing a large watermark or street sign from a cityscape photo where complex geometric lines and shadows exist. Prompt Example: "Erase the red text overlay across the center of the building. Rebuild the brick wall pattern and window frames exactly as they appear in the adjacent sections." Adjustment: For images with strong perspective lines, specify "maintain the vanishing point and linear perspective" to prevent the rebuilt wall from looking distorted.

3. The Portrait Background Softening

Use Case: Cleaning up a caption or social media handle placed over a blurred background (bokeh) in a portrait shot. Prompt Example: "Remove the text overlay at the top of the image. Fill the area with soft, out-of-focus bokeh circles that match the color palette and blur intensity of the background." Adjustment: If the subject is close to the text, add "ensure no artifacts bleed onto the subject's hair or shoulders" to protect the main focal point.

4. The Product Photography Polish

Use Case: Removing a price tag or promotional sticker from a commercial product shot on a plain backdrop. Prompt Example: "Delete the circular price sticker on the product. Reconstruct the solid white background and the subtle shadow cast by the object underneath the sticker." Adjustment: If the product has a glossy surface, include "preserve the specular highlights and reflections" to maintain the realistic lighting of the item.

5. The Document Scan Correction

Use Case: Eliminating a page number or header from a scanned document where the background is paper texture. Prompt Example: "Remove the header text at the top of the page. Generate the clean paper texture with slight grain and uneven lighting to match the rest of the document scan." Adjustment: For documents with faint grid lines or ruled paper, specify "recreate the faint horizontal lines" to ensure the text alignment remains logical after removal.

Optimizing Results and Managing Expectations

While these prompts provide a strong foundation, successful text removal often requires iteration. The model does not guarantee identity or label preservation; instead, it focuses on visual coherence. If the first result shows visible seams or mismatched textures, try refining the prompt to be more descriptive about the specific materials involved.

Remember that Nano Banana 2 supports text-to-image and image-to-image workflows, allowing for flexible experimentation. However, always verify the output against your specific needs. For those looking to explore these capabilities further, Try Nano Banana to access the generator and test these prompts with your own images. By understanding the limitations and strengths of the tool, you can achieve professional-grade edits without needing advanced software skills.