Product Photo Background Contact Shadow Checklist

After changing a product-photo background, review the cutout edge, contact shadow, light direction, and reflections before approving the composite. Compare it side by side with the source photo, because a convincing scene is not proof that the product itself stayed accurate.

This checklist is a suggested review workflow from Kubflow’s practical creative-production perspective. It is not a model provider requirement, and every generated output still needs human review.

Prepare the source files before reviewing the composite

A reliable review starts with an approved reference, not memory. Keep these items together:

• Original product photo: The untouched reference for shape, color, label placement, cap details, openings, and transparent parts.

• New composite: The product placed on its replacement or generated background.

• Background plate: The empty background, when available, so you can identify which shadows and reflections were introduced later.

• Optional edit mask: A mask restricted to the edge, floor, or reflection area that may need revision.

Inspect the composite once at its intended viewing size and again enlarged. Normal size shows whether it works as a commercial image. Enlargement exposes halos, jagged boundaries, clipped details, and unwanted changes to packaging.

Background-removal tools have different capabilities. For example, the BiRefNet Matting documentation describes continuous alpha prediction for smooth transitions around fine structures and semitransparent regions. It also specifies PNG or WebP for transparent output because JPG does not support transparency. That capability does not guarantee that every label, reflective edge, or transparent component will survive accurately.

Use this product photo background contact shadow checklist

1. Check for clean edges

• Look for a fringe carrying the old background color.

• Check curves and diagonal lines for jagged or stair-stepped contours.

• Confirm that corners, handles, straps, openings, and fine details remain intact.

• Inspect glass, clear plastic, glossy metal, and semitransparent materials carefully.

• Compare the silhouette, logo, printed artwork, color blocks, and proportions with the source.

Revision prompt: “Preserve the product silhouette, printed artwork, colors, proportions, and transparent components. Remove the visible edge halo around [specific edge]. Modify only the masked boundary area.”

Use a narrow mask around the defect when your editing method supports masks. Runware’s inpainting guide documents that white mask regions are modified, black regions are preserved, and gray values permit partial modification. It also notes that smaller, well-positioned masks are easier to blend. Preservation is still not guaranteed, so inspect the full product after every edit.

2. Confirm that the product is grounded

• Identify the product’s real contact points, such as a bottle base, shoe sole, or furniture legs.

• Make sure those points meet the depicted floor or surface.

• Reject a visible gap between the object and its shadow.

• Reject a dark outline that follows the entire cutout instead of behaving like a ground shadow.

• Check that the shadow is darkest and tightest at contact, then becomes softer away from the product.

Revision prompt: “Add a subtle contact shadow directly beneath [named contact points]. Keep it attached to the product, darkest at contact, and progressively softer outward. Modify only the masked floor area. Do not alter the product.”

A product can also look pasted on when the shadow is too dark, too wide, or uniformly sharp. The goal is not to add the largest possible shadow. It is to create enough contact information for the product to belong on the surface.

3. Match the product and background lighting

• Find the brightest highlight on the product and infer the apparent key-light direction.

• Compare that direction with the bright side of the background and other visible shadows.

• Confirm that the cast shadow travels away from the apparent light.

• Compare shadow softness with the scene. A broad, soft studio light normally suggests softer transitions than a small, hard source.

• Look for added highlights or color casts that change the product’s approved appearance.

Revision prompt: “Revise only the ground shadow and nearby reflection so they follow the existing camera-left key light. Preserve the product, label, color, proportions, and background.”

4. Review reflections separately

A glossy surface may need both a restrained reflection and a contact shadow. They perform different jobs: the contact shadow anchors the object, while the reflection suggests the surface material.

• Check that the reflection begins at the correct contact point.

• Make sure reflected shapes correspond to the product rather than inventing extra parts.

• Reject unsupported colors, objects, labels, or highlights inside the reflection.

• Reduce reflections that compete with the product or make the scene look artificially mirrored.

Fix the smallest region that can solve the problem

When the product is accurate but the floor treatment is wrong, avoid regenerating the whole image. Mask only the shadow or nearby reflection and describe what must remain unchanged.

Supported GPT Image models can generate images or edit images already present in a Responses API conversation, according to the OpenAI image-generation documentation. Its documented options include output size, quality, format, compression, and background type. The documentation does not promise preservation of product geometry, text, logos, colors, shadows, or reflections.

For that reason, treat “preserve the product” as an editing instruction, not a guarantee. Recheck the entire object after a local revision, even if only the floor was masked.

Worked example: inspect a bottle composite

This example is illustrative. It is not a measured customer result or a claim that the prompt was tested.

Example inputs:

• insulated-bottle-source-front.jpg: Original front-facing photo of a matte navy bottle with a silver cap and white vertical label.

• warm-gray-studio-background.png: Empty warm-gray surface and backdrop with soft illumination from camera left.

• bottle-floor-shadow-mask.png: Black image with a white region limited to the floor immediately beneath and to the right of the bottle.

1. Create a transparent bottle cutout from the source.

2. Place the cutout on the studio background without intentionally regenerating the bottle.

3. If it appears to float, use the floor mask to revise only the grounding area.

4. Compare the result side by side with the original bottle photo.

Copyable example prompt: “Edit only the white masked floor region. Add a subtle contact shadow attached to the bottle’s actual base, with the darkest area at contact and a soft cast shadow extending camera right to match illumination from camera left. Preserve the bottle, silver cap, white label, navy color, silhouette, proportions, and background.”

Decision: reject. Reject the composite if a pale fringe remains around the silver cap or the shadow begins below the bottle with a visible gap. Request a clean cap boundary and a shadow attached to the true base, while leaving the label, shape, and color unchanged.

Apply a clear human-review gate

Approve the composite only when all three acceptance checks pass:

• Clean edges: No obvious halo, jagged contour, clipped feature, or source mismatch.

• Grounded product: Real contact points meet the surface, and the shadow remains attached and plausible.

• Consistent lighting: Highlights, cast-shadow direction, softness, and reflections belong to the same scene.

Also reject any output that changes product geometry, packaging text, logo placement, colors, or essential details. Generation output should not be treated as authoritative evidence of how a product looks.

Kubflow can connect product references, prompts, image steps, and repeatable reruns on a visual workflow canvas. A practical setup is to keep background removal, compositing, local shadow revision, and human approval as distinct steps. For broader review rules, see the guide to what to automate and what to review.

When you are ready to make this inspection process repeatable, build the workflow in Kubflow. Keep final acceptance with a person who can compare every output against the approved source image.