Fix Glowing Halos on Pet Fur in Dark Mode Portraits with Nano Banana

When generating or editing portraits of pets using AI tools, users often encounter a specific visual artifact: a faint, luminous border that appears along the edges of fur against dark backgrounds. This glowing halo effect can make an otherwise high-quality image look artificial or poorly segmented. The issue is particularly noticeable in dark mode portraits where the contrast between the subject and the background is stark. Understanding why this happens and how to correct it within Nano Banana is essential for achieving clean, professional results.

Distinguishing Symptoms from Known Facts

The primary symptom described here is the presence of unwanted edge artifacts, specifically a bright or semi-transparent outline surrounding the pet's silhouette. This is not a defect in the physical camera or the user's monitor but rather a generation artifact common in AI image synthesis when handling complex textures like fur against low-light environments.

It is important to separate plausible causes from verified facts regarding the tool. While some users might suspect their internet connection or browser cache is responsible, the core issue lies in how the model interprets boundary conditions during generation. Verified facts indicate that Nano Banana refers strictly to the AI image generation and editing tool, not any cosmetic brand or physical product. Furthermore, while Google documents specific models like Gemini 3.1 Flash Image under the name Nano Banana 2, the website itself does not guarantee that all features available in the underlying Google documentation are identical or fully exposed on every page of this site. For instance, the existence of a Nano Banana Lite page does not automatically confirm support for multi-turn sequential editing or multiple reference inputs without explicit confirmation.

Therefore, the troubleshooting steps below focus on prompt engineering and workflow adjustments rather than external software fixes. We must rely on the tool's internal logic for contrast and segmentation rather than assuming capabilities that may not be present in every version.

Diagnosing the Root Cause of Edge Artifacts

The glowing halo typically arises from the model struggling to define the precise boundary between the subject and the background. In dark mode scenarios, the lack of ambient light makes it difficult for the algorithm to distinguish fine details like individual whiskers or stray hairs from the shadowy backdrop. Consequently, the model may generate a transitional zone of pixels that appear brighter than the intended background, creating the illusion of a glow.

This problem is often exacerbated by insufficient negative prompting. If the prompt does not explicitly instruct the model to avoid halos or to maintain sharp edges, the default behavior may prioritize smooth blending over crisp definition. Additionally, the choice of model matters. Nano Banana 2 Lite is focused on speed and cost efficiency. It is not optimized for multiple reference inputs or complex multi-turn sequential editing. Using this specific variant for high-detail portrait work requiring precise edge control may yield suboptimal results compared to more robust versions designed for detailed image-to-image workflows.

To diagnose if the issue is prompt-related, try regenerating the image with a modified instruction set. If the halo persists across different seeds and variations, the issue may be inherent to the current model's handling of dark contrasts. However, in most cases, adjusting the textual guidance provided to the generator resolves the artifact.

Fixing the Issue with Refined Prompts and Settings

Resolving the glowing halo requires a strategic approach to prompt construction. The goal is to force the model to recognize the edge as a hard boundary rather than a soft gradient. Start by adding specific negative prompts that target the artifact directly. Include terms such as "no glowing edges," "no halo," "sharp silhouette," and "clean cut" in your negative prompt field. These instructions act as constraints, guiding the model away from generating the luminous border.

Simultaneously, refine the positive prompt to emphasize texture and lighting. Instead of simply saying "pet portrait," specify "high contrast pet portrait," "detailed fur texture," and "defined edges against dark background." This provides the model with clearer context about the desired visual outcome. When working with image-to-image workflows, ensure you are using a version of Nano Banana that supports the necessary complexity for the task. Avoid relying on Nano Banana 2 Lite for this specific fix unless you understand its limitations regarding reference inputs and sequential editing.

Another effective strategy is to adjust the contrast parameters if the interface allows. Increasing the contrast setting can help the model differentiate between the subject and the background more effectively, reducing the need for the model to create a transitional glow. Remember that prompt instructions describe desired outcomes; they do not guarantee identity, label, object, or typography preservation. Therefore, test multiple variations to find the optimal balance between artistic style and technical accuracy.

If you are unsure where to start, the prompt library offers example prompts that users can copy or take into the generator. Look for examples related to portraiture or high-contrast scenes and adapt them to your specific needs. You can also explore the Try Nano Banana link to access the generator and experiment with these settings in real-time.

Verifying the Solution

After applying the refined prompts and adjusted settings, verify the result by comparing the new output with the original problematic image. Check the edges of the pet's fur closely, especially in areas where the fur meets the darkest parts of the background. The glowing halo should be absent, replaced by a clean transition or a sharp silhouette that matches the intended aesthetic.

If the artifact remains, consider iterating on the negative prompts. Try adding stronger descriptors like "black background only" or "no white outlines." It is also worth noting that different models within the Nano Banana family may respond differently to the same prompt. If one version fails to resolve the issue, switching to a more capable model might be necessary, keeping in mind that Google describes Nano Banana 2 Lite as focused on speed and cost, which may limit its ability to handle complex edge cases.

By systematically addressing the prompt structure and understanding the tool's limitations, you can effectively remove edge artifacts and produce high-quality pet portraits without unwanted glowing halos. This process ensures that your images remain professional and visually appealing, regardless of the lighting conditions in the scene.