Nano Banana 2 Image-to-Image: Extending Architecture Vertically with Precision

Understanding Vertical Extension in AI Editing

When working with architectural photography or concept art, the need to modify a structure often arises. You might have a photo of a building that feels too short for its surroundings, or an illustration where a tower needs to reach higher into the sky. The goal is to extend these elements vertically without breaking the laws of physics or the visual style of the original image. This process requires more than just asking the tool to "make it taller." It demands a structured approach to prompt engineering that guides the AI on perspective lines, material continuity, and lighting consistency.

Nano Banana 2 supports text-to-image and image-to-image workflows, allowing users to leverage existing visuals as a foundation. However, the AI does not guarantee identity preservation or perfect structural alignment on its own. Success depends on how clearly you define the desired outcome in your prompt instructions. The system interprets these instructions to generate new content that blends with the input image, but the user must provide specific constraints regarding the extension's geometry and aesthetic.

Structuring Your Prompt for Architectural Integrity

To achieve a seamless vertical extension, your prompt must explicitly describe the relationship between the existing structure and the new addition. Generic requests like "add a roof" often result in disjointed shapes that ignore the building's current angle. Instead, focus on describing the continuation of lines and materials. You should specify the type of extension, such as "add three additional floors" or "extend the spire upward," while emphasizing that the new sections must follow the existing vanishing point and perspective grid.

It is crucial to mention lighting conditions. If the original image has sunlight coming from the left, the new floors must reflect this direction to maintain realism. Describe the texture and color palette of the new elements to ensure they match the base structure. For example, if the building is brick, the added section should be described as "matching red brick facade." Do not assume the AI will automatically infer these details; explicit instruction is necessary to avoid mismatched styles.

Remember that prompt instructions describe desired outcomes but do not guarantee object preservation. The AI may interpret "vertical extension" differently depending on the complexity of the input image. Therefore, using clear, descriptive language about the architectural style—such as "Gothic revival" or "modernist concrete"—helps anchor the generation to the correct aesthetic. These examples serve as templates for structuring your requests effectively.

Step-by-Step Workflow for Vertical Modifications

Executing a vertical extension involves a logical sequence of actions within the Nano Banana 2 interface. Follow these steps to maximize the quality of your results:

1. Select the Correct Model: Navigate to the Nano Banana 2 product page at Try Nano Banana. Ensure you are using the standard Nano Banana 2 model rather than Nano Banana 2 Lite. Google documents Nano Banana 2 Lite as focused on speed and cost, noting it is not optimized for multiple reference inputs or complex multi-turn sequential editing. For architectural tasks requiring precision, the standard model is the appropriate choice.
2. Upload Your Base Image: Start by uploading a high-quality image of the building you wish to modify. Ensure the image clearly shows the top edge of the structure and the surrounding context to help the AI understand the horizon line and perspective.
3. Craft the Detailed Prompt: Enter your prompt focusing on the specific extension. Use phrases like "extend the building upwards by adding two floors" and include details about materials and lighting. Label any untested variations as examples to manage expectations.
4. Generate and Review: Submit the request and review the output. Check if the new elements align with the original perspective lines. Look for inconsistencies in window patterns or rooflines that suggest a lack of structural logic.
5. Iterate if Necessary: If the first result looks disconnected, refine your prompt. Add more specific constraints about the angle or the connection point between the old and new sections. Avoid repeating the same paragraph or generic advice; instead, tweak the specific descriptors of the architecture.

Evaluating Results and Troubleshooting Common Issues

Judging the success of a vertical extension requires a critical eye for detail. A successful edit will show continuous lines where the walls meet the new floors, consistent shadowing across all levels, and a unified color temperature. If the new section appears to float above the building or has a different texture, the prompt likely lacked sufficient detail regarding material continuity.

Common issues include distorted windows or misaligned rooflines. To fix this, try simplifying the prompt to focus solely on the extension geometry before adding stylistic details. Another frequent problem is the AI adding unrelated objects like trees or clouds where the building should extend. In such cases, reinforce the negative space requirements in your prompt, specifying that the area above the building should remain empty or contain only sky.

While Nano Banana 2 offers powerful capabilities, it is important to remember that no AI tool guarantees perfect outcomes. The generated images are interpretations based on your input. By carefully structuring your prompts and selecting the right model version, you can significantly improve the likelihood of achieving a professional-looking architectural modification. Always refer to the official documentation for the latest updates on model capabilities and limitations.