40 essential terms for cross-border e-commerce sellers using AI tools — from Amazon main image policy to virtual try-on. Every entry links to the AvoAgent tool that covers it.
This glossary defines the core vocabulary of AI-assisted e-commerce operations — product photo processing, listing translation, marketplace compliance, and generative AI techniques. Each term is written for non-technical sellers, with a link to the AvoAgent tool where that concept applies. Use it as a reference when you're evaluating tools, writing SOPs for your team, or understanding what specific marketplace rules mean in practice.
The process of translating product titles, descriptions, and attribute lists into a target market's language while preserving brand names, model numbers, and marketplace-specific phrasing. Different from generic machine translation because it keeps commerce-specific terminology intact.
Translation that keeps proprietary terms (brand names, model numbers, protected phrases) untranslated while localizing the surrounding descriptive text. Critical for e-commerce because "AirPods Pro" should not become "空气吊舱专业版" in Chinese listings.
AI-based translation of text embedded inside a product image (infographic callouts, packaging labels, spec highlights) without redesigning the image. Detects text regions, translates, and re-renders in target language.
Localization of Amazon A+ Content (Enhanced Brand Content) — multi-image marketing modules on brand-registered listings. Requires typography-preserving image translation because A+ modules are typography-heavy branded designs.
AI-based removal of the background around a product's main subject. Produces either a transparent PNG (for compositing onto other scenes) or a pure white background (marketplace hero image standard). Also called "background removal" or "product cutout".
Image file format supporting alpha channel — pixels can be fully transparent, letting the product photo overlay any background. Common in cross-border e-commerce for lifestyle compositing and multi-marketplace listings.
AI-based erase of supplier watermarks, store logos, price stickers, and promotional overlays from product photos. Uses inpainting to fill the background where the watermark was. Essential for marketplace-compliant main images.
AI technique that fills a masked area of an image with plausible content, matching surrounding textures and colors. Used to reconstruct the background after removing watermarks, text, logos, or unwanted objects from a product photo.
AI-based enlargement of a low-resolution image (2×, 3×, or 4×) that reconstructs plausible detail rather than blurry pixel stretching. Also called super-resolution. Used to bring supplier photos up to marketplace minimum-resolution requirements.
AI technique that infers a high-resolution version of a low-resolution image by predicting plausible detail. The technical name for AI image upscaling. Cannot recover information that was never captured in the original.
The proportion between an image's width and height (1:1 square, 3:4 vertical, 4:5, 16:9 banner). Different marketplaces require different ratios: Amazon prefers 1:1, TikTok Shop favors 3:4 vertical, Shopify feeds use 4:5 or 1:1.
AI-based reframing of a product photo to a target aspect ratio with the subject auto-centered. Different from a simple center-crop because it detects the main subject and keeps it in frame regardless of its original position.
AI-based scan of a product image to report the presence and location of text regions, watermarks, logos, QR codes, and other embedded elements. Used as a pre-processing step before batch photo cleanup workflows.
AI-based localization of a model's skin tone in an existing apparel photo to match a different regional target market. Keeps the garment, pose, lighting, and composition identical — only the model's skin tone changes.
AI-based generation of an on-model apparel photo from a flat-lay garment shot plus a model reference photo. Replaces physical apparel photoshoots by inferring how the garment would appear when worn.
Applying the same tool operation to multiple images or listings in a single run, rather than processing each one individually. AvoAgent supports batch mode for up to 30 images per run in Cutout, Smart Erase, Smart Crop, and HD Upscale.
The primary product photo shown as the main image in a marketplace listing. Amazon calls it the "main image" and requires a pure-white background with no text or logos. Hero images have the highest impact on click-through rate from search results.
Amazon's fulfillment service where sellers ship inventory to Amazon warehouses and Amazon handles storage, packing, shipping, and customer service. FBA sellers can list on Amazon marketplaces globally with local inventory.
Amazon's program for verified brand owners that unlocks A+ Content (Enhanced Brand Content), Amazon Stores, brand analytics, and enhanced counterfeit protection. Requires a registered trademark. Important for image translation quality standards.
Amazon's Enhanced Brand Content — multi-module image and text sections on brand-registered product pages. Typography-heavy, brand-designed, and requires high-fidelity translation across localized marketplaces.
Shopify's enterprise plan for high-volume merchants ($2000+/month). Offers advanced customization, dedicated support, and higher throughput. Shopify Plus stores typically manage 300-1000+ SKUs and heavily use bulk photo prep tools.
TikTok's integrated e-commerce marketplace where sellers list products discoverable through videos, live streams, and the Shop tab. Uses 3:4 vertical product photos, video-first product discovery, and regional marketplaces (UK, US, EU, SEA).
One of the fastest-growing global marketplaces (2022+), owned by PDD Holdings. Strict listing formatting rules — character-limited titles, banned keywords per category, and required attribute keywords. High SKU volume rewarded by the algorithm.
Alibaba's cross-border B2C marketplace, widely used as a supplier source by Shopify dropshippers. AliExpress product photos typically require watermark removal, background cutout, and description rewriting before being usable on a Western storefront.
Selling products from one country to buyers in another country through international marketplaces or direct-to-consumer stores. Requires translation, currency handling, and per-region compliance for images and listings.
E-commerce model where the seller does not hold inventory. When a customer orders, the seller purchases from a wholesaler (typically AliExpress or 1688) who ships directly to the customer. Requires bulk photo cleanup from supplier feeds.
E-commerce strategy where a seller creates their own brand name on generic products manufactured by a third party. Common on Amazon FBA. Requires unique product titles, brand-consistent photos, and brand registry for A+ Content.
A unique identifier for each distinct product variant a seller offers. Different sizes, colors, or bundles of the same product are separate SKUs. Sellers with 100+ SKUs benefit most from bulk photo processing and title generation.
The practice of improving marketplace product listings — titles, descriptions, images, keywords — to improve search ranking, click-through rate, and conversion. Combines SEO principles with marketplace-specific algorithm knowledge.
The featured seller position on an Amazon product page where the "Add to Cart" button routes to a specific seller. Winning the Buy Box requires competitive pricing, high-quality photos, and strong seller performance metrics.
Amazon's requirement that main product images have a pure white (RGB 255,255,255) background, no text, no logos, no watermarks, and the product filling 85%+ of the frame. Non-compliant main images can cause listings to be suppressed.
Marketplace-required minimum pixels on the longest side of a product image — Amazon requires 1000px+ for zoom, and recommends 1600px+ for optimal quality. Below the minimum, listings lose zoom function and appear low-quality in mobile.
Adherence to a specific e-commerce platform's rules for product photos, titles, descriptions, and category attributes. Non-compliant listings get suppressed, delisted, or flagged for review. Each marketplace has different rules.
Words that marketplaces prohibit in product titles or descriptions — subjective claims ("best", "cheapest"), medical claims ("cures", "treats"), trademark violations, or restricted category-specific terms. Automated title generators check against banned lists.
Modern translation approach using deep neural networks trained on billions of sentence pairs. Handles context better than earlier statistical machine translation. AvoAgent's Text Translation is NMT tuned for e-commerce.
The AI field dealing with image understanding — object detection, segmentation, generation, and transformation. Underlies every AvoAgent image tool: Element Detection identifies elements, Cutout segments the subject, Smart Erase inpaints backgrounds.
AI models that create new content — text, images, video — rather than only classifying or transforming existing content. Title Generator and Description Generator are generative AI. Virtual Try-On is generative image AI.
Job model where a user submits a request and receives the result later (seconds to minutes) rather than immediately. AvoAgent's premium tools (Virtual Try-On, Model Skin Replacement, Image Translation Pro) run async because they involve heavier AI computation.
Pricing model where users pay only for the specific jobs they run, with no monthly subscription or minimum commitment. AvoAgent charges per successful job — failed calls are auto-refunded. Also called consumption pricing or pay-as-you-go.
AvoAgent's policy of automatically refunding credits back to the account balance when a tool call fails (model timeout, low-quality result, unsupported input). No manual refund request needed. Ensures users only pay for successful outputs.
Sign up and get free trial credits. Test every tool the glossary references — no card needed.