Improving Image Quality

Virtual try-on results are only as good as the input images. If your outputs look distorted, unrealistic, or low-quality, these guidelines will help you get significantly better results.

Garment Photo Best Practices

The garment image is the foundation of every virtual try-on. Getting this right has the biggest impact on output quality.

Resolution — Use images at least 1024 x 1024 pixels. The minimum supported resolution is 512 x 512, but lower resolutions produce noticeably softer results with less fabric detail. If your product photos are smaller than 1024px, consider re-shooting or using your original high-resolution files rather than web-optimized versions.

Background — Use a plain, neutral background (white, light gray, or solid color). Busy backgrounds confuse the garment extraction step, leading to artifacts around edges. If your existing product photos have complex backgrounds, the AI will still work, but results improve dramatically with clean backgrounds.

Garment visibility — The entire garment must be visible in the frame. Common problems:

• Sleeves cropped at the edge of the image
• Hemline cut off at the bottom
• Collar hidden by styling or folding

Flat-lay photography or mannequin shots provide the most consistent results because the full shape of the garment is visible without distortion from a model’s pose.

Lighting — Even, diffused lighting without harsh shadows. Side lighting creates shadow patterns that the AI may interpret as part of the garment’s design. Natural daylight or softbox studio lighting works best.

Customer Photo Guidelines

When customers submit photos for virtual try-on (via Instagram DM or your storefront), better inputs lead to better results. Share these guidelines with your customers:

Pose — Face the camera directly with arms slightly away from the body. The AI needs to see the torso clearly to map the garment correctly. Arms crossed, hands in pockets, or turned poses reduce accuracy.

Framing — For tops, the photo should show at least from the waist up. For dresses or full outfits, a full-body photo is required. The closer the framing matches the garment type, the better the result.

Lighting — Well-lit photos with even lighting produce the most realistic outputs. Common issues:

• Backlit photos (bright window behind the person) create silhouettes that lose body detail
• Harsh overhead light creates dark shadows under the chin and arms
• Very dim photos lack the detail the AI needs to accurately map proportions

Background — While not as critical as garment photos, a clean background helps. Busy environments with other people, mirrors, or patterned walls can confuse the AI.

Common Issues and Fixes

Blurry or soft output — Almost always caused by low-resolution input images. Check that both the garment and customer photos are at least 512 x 512 pixels, ideally 1024 x 1024 or higher.

Garment edges look jagged or unnatural — The garment photo likely has a complex background. Re-shoot on a plain background or manually crop the garment before uploading.

Wrong proportions or distorted fit — The customer photo may not show enough of the body, or the person’s pose makes it difficult to map the garment. Ask for a front-facing photo with arms visible.

Color looks different from the original — Lighting differences between the garment photo and customer photo can cause color shifts. Using consistent, neutral lighting in garment photography minimizes this.

Parts of the garment are missing — Ensure the full garment is visible in the product photo. If sleeves or hemlines extend beyond the frame, the AI can’t reconstruct what it can’t see.

Batch Processing Tips

If you’re running virtual try-on across your entire catalog:

1. Standardize your product photography — Same background, same lighting, same framing for every garment. This creates consistency across all try-on outputs.
2. Review the first 10-20 results before processing the full catalog. Catching issues early saves credits and time.
3. Separate by garment type — Tops, dresses, and outerwear may need different photo angles. Group similar items and optimize settings for each batch.
4. Use the highest resolution available — Pull from your original photography files, not compressed web images or thumbnails.