Nano Banana 2: Recolor Garments Without Losing Fabric Texture

Why Preserving Texture Matters in AI Fashion Editing

When editing fashion imagery, the most common pitfall is losing the tactile quality of the fabric. Users often request a simple color swap, only to receive an image where the material looks like flat plastic or melted wax. The specific challenge addressed here is changing the clothing color on a model without altering the fabric texture. This requires a prompt strategy that explicitly separates the attribute of color from the physical properties of the weave, stitching, and light interaction.

Nano Banana 2 supports text-to-image and image-to-image workflows, making it suitable for this type of precise manipulation. However, the tool does not guarantee identity, label, object, or typography preservation in every instance. Therefore, crafting the right instruction is essential to guide the model toward maintaining the intricate details of the original material while shifting the hue. This approach ensures that a denim jacket remains visibly denim, even if the shade changes from indigo to black.

Core Prompt Strategies for Color Swaps

To achieve a successful recoloring task, your prompt must be descriptive about the desired outcome while strictly defining what must remain unchanged. Below are five materially different usable prompts designed for various scenarios. These are examples of how to structure instructions; they do not guarantee identical results across all generations.

1. The Direct Material Lock Prompt: "Change the color of the red silk dress to emerald green. Keep the exact same silk weave pattern, sheen, and lighting folds exactly as they appear in the source image. Do not smooth the texture or alter the drape." Use Case: Best for high-end fashion photography where the specific sheen of silk or satin is critical. It works when the user wants a direct substitution without any stylistic drift. Adjustment: If the result looks too matte, add keywords like 'high gloss' or 'reflective surface' to the prompt.

2. The Lighting Consistency Focus Prompt: "Recolor the blue wool coat to charcoal gray. Maintain the original shadow depth, highlight placement, and the rough texture of the wool fibers. Ensure the new color interacts with the existing light sources exactly as the blue did." Use Case: Ideal for outdoor shots or studio lighting where shadows define the shape of the garment. This helps prevent the new color from looking like a sticker pasted over the image. Adjustment: If the lighting looks inconsistent, specify the light direction (e.g., 'softbox lighting from left') in the prompt.

3. The Weave Pattern Preservation Prompt: "Transform the plaid shirt from orange to navy blue. Strictly preserve the grid lines of the plaid pattern and the visible cotton thread texture. Do not blur the pattern or change the fabric weight appearance." Use Case: Essential for patterned garments where the geometry of the design is as important as the color. This prevents the AI from smoothing out the checks or stripes. Adjustment: For complex patterns, increase the emphasis on 'pattern integrity' or 'sharp edges' in the instruction.

4. The Realistic Wear and Tear Prompt: "Change the color of the distressed black jeans to faded indigo. Retain all existing rips, fraying edges, and crease marks. The new color should look naturally dyed into the worn fabric, not painted on top." Use Case: Perfect for streetwear or vintage styles where imperfections are key features. This ensures the AI doesn't 'heal' the tears while changing the color. Adjustment: If the rips disappear, explicitly state 'preserve all damage and distress marks' at the beginning of the prompt.

5. The Multi-Material Distinction Prompt: "Recolor the leather jacket sleeves to burgundy and the cotton body to cream. Keep the grain of the leather and the softness of the cotton distinct. Do not mix the textures or make them look like the same material." Use Case: Useful for outfits combining different fabrics. This helps the model understand that different parts of the image require different texture rules. Adjustment: If textures blend, separate the instructions by section, specifying 'leather texture' and 'cotton texture' for each area individually.

Model Selection and Workflow Limitations

While Nano Banana 2 is capable of these detailed edits, users should be aware of the distinctions between available models. Google documents Nano Banana 2 as Gemini 3.1 Flash Image, while Nano Banana Pro corresponds to Gemini 3 Pro Image. There is also a Nano Banana 2 Lite version, identified as Gemini 3.1 Flash Lite Image, which is focused on speed and cost.

It is crucial to note that Nano Banana 2 Lite is not optimized for multiple reference inputs or multi-turn sequential editing. If your workflow requires refining a color change through several steps or using multiple reference images to ensure texture accuracy, you should avoid the Lite version. Instead, utilize the standard Nano Banana 2 capabilities for better fidelity.

Remember that prompt instructions describe desired outcomes but do not guarantee identity or object preservation. Always treat generated results as examples to be refined. For more information on the capabilities of these tools, visit Try Nano Banana. By carefully constructing your prompts to isolate color from texture, you can achieve professional-grade fashion edits that respect the physical reality of the materials depicted.