Product photo background replacement in ComfyUI: define what may change before you batch
Online-store product images can have their backgrounds replaced in ComfyUI, but the product must remain faithful to the physical item. This guide covers masking decisions, batch submission through the API, marketplace requirement checks and the accuracy risk of altered images.
Purpose and scope
This guide covers replacing or extending the background of online-store product photographs in ComfyUI while the product stays faithful to the physical item.
Source-supported context
ComfyUI's inpainting tutorial describes regenerating only a masked region, with a grow-mask setting that creates a transition area around the mask edge (Source 2). The ComfyUI server accepts workflows posted to /prompt, which returns a prompt_id, and exposes /history, /view and /upload/image (Source 1). Shopify lists a 20 MB limit, 25 MP maximum, and JPEG, PNG, WEBP, HEIC and GIF formats (Source 4). Amazon's Seller Central product image guide (US marketplace view) requires a pure white main-image background (RGB 255, 255, 255), asks that the product be shown as 85% of the image, sets 500 pixels minimum and 10,000 maximum on the longest side, recommends 1,000+ pixels to enable zoom, and requires that images represent the product's real scale, quantity and color (Source 3). It also bars text or pricing details (swatch image only), review or claim imagery and Amazon logos or badges, and says non-compliant images may be removed (Source 3). For photorealistic AI-generated people, the same page says some jurisdictions may require disclosure and asks sellers to add the keyword contains-synthetic-performer to the image's dc:subject (XMP) field (Source 3). The FTC's 1983 Policy Statement on Deception lists three elements: a representation, omission or practice likely to mislead the consumer; judged from the perspective of a consumer acting reasonably; and material, meaning likely to affect the consumer's conduct or decision (Source 5).
Before you start
1. Keep each SKU's original photograph and a colour and label reference.
2. Open each channel's current image rules.
3. Have a saved workflow and model records.
4. Name a reviewer with access to the physical product.
Steps and checks
The sequence below is an editorial procedure based on the cited context, not a diagnosis already confirmed for your environment.
Step 1. Before generating, list per SKU the exact label text, logo placement, reference colours and proportions a buyer relies on. This list is the acceptance standard.
Step 2. Decide which pixels may change: keep the product pixels from the photograph and regenerate only the surroundings through a mask aligned to the product edge, with a controlled grow value (Source 2).
Step 3. Fix the catalogue look once: canvas, margin, background, shadow and crop rule. Reuse one saved workflow and the same recorded model files for every SKU.
Step 4. For batches, export the workflow in API format (see the linked guide on native API submission), submit one prompt per SKU to /prompt, keep the returned prompt_id against the SKU, and fetch outputs from /history and /view (Source 1). The routes page also lists /upload/image for uploading source photographs.
Step 5. Compare every output with the source at full size, not as a thumbnail: read label text, check logo shape, sample colours, and check that no edge, port or accessory was added or reshaped. Reject failures.
Step 6. Check channel rules on publication day: background, size, format, fill, overlays and, if the photo shows people, AI-disclosure metadata (Source 3) and, for Shopify, file size, megapixels and format (Source 4).
Step 7. Archive each approved image with its source photo, workflow, model identities, seed, reviewer and date.
Completion check
Each image has a source photo, a passed full-size fidelity check, a dated channel check and an archived workflow, model and seed record.
Limitations and cautions
Re-encoding may alter small text, logos and shapes outside the mask; this is untested here and not quantified in the cited pages. The Amazon figures above are from the US Seller Central help page as displayed on 2026-10-06; that page is rendered by JavaScript and its requirements can differ by category and marketplace, and it points to separate category style guides that were not read. An older Amazon-staff post in the Seller Central forum (Source 7) lists somewhat different figures (85 to 100 percent fill, 1,000 pixel minimum, no text, logos or watermarks on main images); it is a community post, so treat the help page as authoritative and confirm current rules in Seller Central. Shopify's product-media page defers specifications to its file documentation (Source 6). Rules vary. Whether an altered image misleads buyers depends on your jurisdiction and contracts; this is not legal advice, and the FTC statement is US-only.
Original sources
- ComfyUI server routes — checked 2026-10-06.
- ComfyUI inpainting — checked 2026-10-06.
- Amazon Seller Central product image guide — checked 2026-10-06.
- Shopify file requirements — checked 2026-10-06.
- FTC deception policy — checked 2026-10-06.
- Shopify product media — checked 2026-10-06.
- Amazon Seller Central forum post (Amazon staff, community) — checked 2026-10-06.
Prepared 2026-10-06 from the sources above. No GPU run, quality measurement or runtime guarantee is supplied.
Related troubleshooting and guides
Was this page helpful?
Anonymous. Only a yes/no counter is stored; no account, IP address or device details.
Sources & references
Editorial research from the cited public sources, dated 2026-10-06. This site did not run or test the workflow described.
01ComfyUI server routesSource checked: 2026-10-0602ComfyUI inpainting tutorialSource checked: 2026-10-0603Amazon Seller Central product image guideSource checked: 2026-10-0604Shopify file requirementsSource checked: 2026-10-0605FTC Policy Statement on DeceptionSource checked: 2026-10-0606Shopify product mediaSource checked: 2026-10-0607Amazon Seller Central forum: product photography standards (Amazon staff post)Source checked: 2026-10-06Report an issue · af5ef169-61c8-5fb9-967c-342cf49a9f5e
Related reading
Pages linked to this one by the editors.
A painted mask is only part of an inpainting workflow. Check the image, mask interpretation and supported model/encoding path together.
This guide concerns an instance you own or are authorized to use. ComfyUI's native HTTP API is distinct from this site's knowledge MCP and does not authorize access to third-party services.
After execution, identify which nodes produced results and where those results reside. Node IDs, prompt IDs and filenames are different identifiers.
A model name can cover bundled checkpoints, split weights, quantized variants, text encoders and VAEs. Preserve enough identity information to know which artifact a workflow actually used.