Amazon MAIN (single, facing left, 45°)
Derived from the real capture at the exact angle convention. The suppression-proof slot.
Footwear product photography: the single-shoe 45° cover, pair discipline and a sole tread you didn't invent
Footwear has the most specific main-image rule in commerce, and it's a photograph rule: Amazon requires the MAIN image to show a single shoe, toe facing left, at a 3/4 view (their style guides say ~40°, G1881 says 45°), pure white RGB 255, product filling ≥85% of the frame — no pair, no right shoe, no model, no box. Get the angle, handedness or fill wrong and the ASIN suppresses from search. 'Shoe product photography' is ~140 US searches at difficulty 7. Meanwhile the AI tells for shoes are brutal and well-known: mismatched pairs, tread patterns redrawn into generic zigzags, swoosh-adjacent geometry that invites trademark trouble. Berrys keeps the silhouette, tread and material grain locked to real capture and does the volume work around it — per-marketplace crops, pair variants, on-foot context and size charts.
Reviewed October 2026 · Prepared by Berrys · Technical specifications & Canvas node workflows
Footwear image rules by channel (checked 22 Oct 2026)
| Parameter / Slot | Marketplace & Production Requirement |
|---|---|
| Amazon MAIN | Single shoe (UK guide: left shoe), toe facing left, 3/4 view ~40–45° from slightly above · pure white #FFFFFF · ≥85% fill · flat-lay shot convention |
| Amazon prohibitions | No pairs in MAIN, no right shoe alone, no models, no boxes/accessories, no text or logos; children's footwear has additional restrictions |
| Zoom spec | ≥1,000 px long side for US zoom (UK guide: 1,600+ px), JPEG preferred, RGB, sharp and fully in focus |
| Amazon alternate slots | Pair, heels, tread/sole, side profiles, detail macro — variant codes MAIN/PT/FRNT/BOTTOM etc. |
| Shopee | 1:1 squares, cover on solid bg with ≥60% fill, no text on cover; up to 9 slots; apparel/footwear sub-categories require a separate size-chart image |
| Size chart | Generate from your last's actual mm/cm measurements (foot length → size grid); a chart copied from a generic template is a return factory |
| On-foot / lifestyle | Allowed in secondary slots across channels; keep the shoe pixels locked to the real upper, midsole and outsole |
| Video | Step-in / walk cycles from the locked reference; outsole flex and tread contact must match the real construction |
Use all 9 slots: a Berrys gallery plan
Pair, facing left
Secondary per Amazon; cover-eligible elsewhere. Symmetry checked — no mismatched eyelets.
Heel & medial side
The two angles buyers use to sanity-check silhouette honesty.
Tread macro
Outsole pattern cropped from real capture — a redrawn tread is a different shoe.
Upper/material detail
Knit grain, leather pebble, mesh weave at zoom; texture derived, never imagined.
Size chart
Real mm grid per the last; Shopee apparel/footwear requires the separate chart image.
On-foot context
Street/trail/gym branch; garment and ground vary, shoe geometry stays locked.
What's in the box
Shoe + extras shot that prevents missing-accessory disputes.
Walk-cycle video
Step-in loop from the locked reference, 9:16 for slots and ads.
Where generic AI breaks footwear listings
Pairs that aren't twins
Models generate the left and right shoe as separate inventions — eyelet counts, panel stitching and swoosh curvature diverge, and sneaker communities screenshot mismatched pairs as the canonical AI tell. Pair discipline starts from one locked geometry and mirrors honestly.
Tread redraws = trademark + function
Outsole patterns are functional IP (and sometimes protected trade dress). A model 'improving' a tread creates a shoe that doesn't ship — and can create one that can't be sold. Sole imagery crops from the real capture.
The 45° convention is unforgiving
Facing-right MAINs and pair MAINs are auto-suppressed; a beautiful render at the wrong angle is a delisted ASIN. The export graph bakes the handedness/angle rule per channel instead of hoping the upload passes.
How Berrys produces the set
Reference-first pipeline
One true-angle studio (or approved factory) capture per model/colorway seeds everything downstream.
Colorway fan
Per-colorway File node references keep upper/material truth across the family — no cherry-red panel on the navy pair.
Channel presets
Amazon's 9-slot plan and Shopee's size-chart slot are graph outputs, so the same set serves US and SEA without a re-shoot.
Production workflows in Berrys Canvas
Single-shoe 45° cover → 9-slot marketplace suite
1. File Node: Upload raw photograph of single shoe (toe facing left at 45° 3/4 view).
2. Brand System Node: Enforce Amazon footwear MAIN rule (single shoe, toe left, pure white #FFFFFF, ≥85% fill, no box or pair).
3. Image Node (Nano Banana 2.1): Render 2048px compliant hero packshot.
4. Fan-out: Generate matched pair view, heel/medial profile, real outsole tread macro, upper material zoom, and on-foot lifestyle context.
Multi-colorway family batch run
1. File Node: Ingest silhouette master and colorway swatch cards.
2. Brand System Node: Lock upper mesh, leather pebble grain, and midsole texture physics.
3. Image Node (Nano Banana 2.1): Batch render all colorway variants under identical studio lighting.
Step-in walk-cycle video slot
1. Image Node: Connect verified on-foot still master.
2. Video Node (Omni Flash / Veo 3.1): Render 6-second natural step-in and ground contact motion clip.
3. Output Gallery: Export 9:16 for TikTok/Reels and 1:1 for PDP video slot.
Berrys Plugins · Model Context Protocol (MCP)
Create in chat. Keep it on your canvas.
Already working in an AI agent? Connect Berrys to Claude, Cursor, ChatGPT, Codex, or Hermes. Your agent plans the footwear suite and runs the shots on your Berrys canvas, with saved Brand Memory and credit approval in chat.
Connector URL:
https://app.berrys.io/api/mcpRemote MCPDrop-in Agent Brief (Footwear)
Using Berrys, create an Amazon & Shopee compliant 9-slot footwear suite for my sneaker reference in ./assets: 1. File node: Ingest single shoe reference photo (left shoe, toe facing left at 45° angle). 2. Brand System node: Set pure white background (#FFFFFF). Lock product bounding box to 85% frame fill. Ensure outsole tread pattern and eyelet geometry remain strictly locked to the physical sample. 3. Image node (Nano Banana 2.1): Render 2048px Amazon-compliant MAIN cover. 4. Secondary images: Generate matched pair view, heel profile, close-up tread macro, and upper knit/leather texture zoom. Estimate credit cost and request approval before generating.
Compliance & production boundaries
Footwear renders live or die on geometry the buyer can verify: tread, eyelets, swoosh curvature, handedness, mm sizing. This workflow automates the nine-slot machine around a truthful capture and enforces each marketplace's photo-law (single-left-45°, size-chart slot, cover fill). Sizing performance, cushioning feel and durability claims stay with the physical shoe — no image layer can honestly assert them.
When NOT to use AI generation for this
- Pre-order renders of shoes that don't physically exist yet (label as representations at best)
- Tread/logo geometry that differs from the manufactured item (functional IP + listing-truth)
- Fit, width or performance claims a sample hasn't confirmed
- Sports-safety contexts implying tested protection (ankle support, grip ratings)
Frequently asked questions
Why does Amazon insist on a single shoe facing left?
It's their documented style-guide convention for footwear MAIN images (left shoe, toe pointing left, ~40–45° 3/4 view on pure white) — consistent search results. Non-compliant MAINs are suppressed from search, which costs the listing all organic visibility.
Can AI generate a shoe that doesn't exist yet for pre-orders?
You can, and buyers will treat the discrepancy as fraud when it ships. Pre-launch renders need 'image is a representation' labeling at minimum; the safe pipeline is approved-sample photography as the lock.
Does the size-chart image really need the real last measurements?
Yes — chart-to-product mismatch is the number-one footwear return driver on SEA marketplaces. Shopee now requires the separate chart slot in many apparel/footwear categories, so a generated generic chart compounds the problem.
Can the canvas do on-foot shots without a model shoot?
Context/worn scenes branch from the locked shoe; the shoe pixels carry real capture truth. Where fit itself is the claim (sizing advice, width promises), a physical sample still rules.
Related tool comparisons & guides
In-Depth Guides & Research from Berrys Journal
Create on-brand assets without booking a studio
Start free on Berrys Canvas — 50 credits, no card required — or send Berrys Studio a brief and get finished campaign visuals with agency art direction.