White-bg 12-o'clock master
Straight-on cover, dial crisp, reflection structure from the real capture.
Watch product photography: dial legibility, bracelet reflections and lume — without the caseback confabulation
The AI-image community's favorite 'gotcha' is a generated watch: wrong subdial count, indices that melt into the chapter ring, a crown at 4 o'clock on a 3-o'clock case, mirror bracelet holding a reflection of a room that doesn't exist. 'Watch product photography' is only ~90 US searches a month but competition sits at 3 — near-uncontested for anyone who can publish the category's hard craft honestly. Watches are also the highest AOV-to-image-ratio purchase in ecommerce: buyers study a 200 px thumbnail like appraisers. Berrys derives the full marketplace set from real macro capture with reflection-controlled lighting, keeping dial typography, bezel graduations and case geometry physically true across every crop.
Reviewed October 2026 · Prepared by Berrys · Technical specifications & Canvas node workflows
Watch image requirements & craft rules (checked 22 Oct 2026)
| Parameter / Slot | Marketplace & Production Requirement |
|---|---|
| Marketplace cover | Amazon: pure white, ≥85% fill, single item, no model, no box · Shopee: solid bg ≥60% fill, 1:1, ≤2 MB |
| Dial legibility | Applied indices, lume plots and chapter-ring numerals must be crisp at 1,000+ px zoom — the dial is the product identity |
| Reflection control | Bracelet/crystal is a mirror: capture with gradient controls (dark-field/bright-field) and lock that highlight structure; generation must not invent reflected scenes |
| Geometry truth | Subdial count, hand style, crown position, bezel graduations, caseback engraving — every generation must match the reference exactly; these are the tells buyers screenshot |
| Scale | Case diameter in mm on a size card (your spec sheet), lug-to-lug, thickness — watch buyers return on millimeters |
| Lume shot | Dark-glow exposure derived from a real long-exposure capture, matched to the actual compound's color and coverage |
| Video | 360° spin + wrist-sweep; motion reveals geometry errors that stills hide — only from locked references |
Use all 9 slots: a Berrys gallery plan
45° case angle
Shows thickness, crown and lug profile — the angle that proves it's the same watch.
Dial macro
Index/lume/hands detail at zoom; typography preserved, never re-drawn.
Bracelet/clasp
Link finish and clasp type; buyers of divers care more about the clasp than the bezel.
Size card
mm scale + lug-to-lug from the spec sheet — the slot that prevents the 'chunkier than expected' return.
Wrist context
On-wrist branch from the locked piece; skin/lighting vary, watch pixels don't.
Where generic AI breaks watch listings
Mirrors invent rooms
A polished bracelet reflects a studio that generative models never had. Hallucinated reflections carrying text, hands or logos are the signature AI artifact — and watch buyers are the most artifact-literate shoppers alive. Derived from controlled real captures, reflection structure is physics, not prompt.
Dials are documents
Brand text, model names, 'SWISS MADE' placement and graduations are small typographic facts. Models blur and re-spell them. The workflow treats the dial as a locked texture region cropped from capture — legibility at zoom is non-negotiable.
Complication confabulation
Chronographs get extra subdials, date windows drift position, bezels lose clicks. Every generation must pass the geometry checklist (subdials/crown/hand-style/bezel) before it ships; the reference lock is what makes that checkable.
How Berrys produces the set
One macro session per reference
Real capture seeds the family; the graph scales it to every marketplace crop, context shot and spin.
Collection consistency
Same lighting grammar across a drop so the storefront reads as one atelier, not a grey-market tray.
Motion nodes
Spin/wrist clips from the locked reference — the video layer that converts at $200+ AOV without a turntable rig per SKU.
Production workflows in Berrys Canvas
Macro stack master → 6-slot horology listing
1. File Node: Upload focus-stacked macro capture of watch head and dial.
2. Brand System Node: Lock dial typography, subdial geometry, hand style, and sapphire crystal anti-reflective coating reflections.
3. Image Node (Nano Banana 2.1): Render 2048px 12-o'clock front cover with pure white background and gradient-controlled bracelet reflections.
4. Fan-out: Generate 45° case profile (showing crown and lug thickness), 2048px dial macro, clasp detail, and mm lug-to-lug size card.
Collection drop-day batch run
1. File Node: Connect reference captures for all dial/case variants.
2. Brand System Node: Enforce unified horology bright-field studio lighting.
3. Image Node (Nano Banana 2.1): Batch render entire collection ensuring zero complication confabulation across models.
Wrist-sweep and 360° spin video cut
1. Image Node: Connect verified watch hero master.
2. Video Node (Omni Flash / Kling 4.0): Render 6-second smooth wrist orbit showcasing bracelet link articulation and dial light play.
3. Output Gallery: Export 1440×2560 vertical video for Reels and square for PDP.
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 watches & horology 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 (Watches & horology)
Using Berrys, create a 6-slot horology watch PDP suite from my reference in ./assets: 1. File node: Ingest focus-stacked macro photo of watch head. 2. Brand System node: Lock dial typography, subdial count, hand geometry, and crown position as immutable zones. Apply 'Horology Studio' dark-field/bright-field reflection lighting preset. 3. Image node (Nano Banana 2.1): Render 2048px straight-on 12 o'clock hero packshot on pure white. 4. Macro pass: Render 2048px extreme zoom on dial indices, logo printing, and date window. 5. Case profile: Render 45° side view showcasing case thickness, lug contour, and crown finishing. Estimate credit cost and request approval before generating.
Compliance & production boundaries
A watch image is close to a certificate: buyers cross-check the render against the reference number's known geometry. This workflow gives you the full marketplace set, context imagery and motion at catalog speed — from captures that carry the truth. Where the listing makes mechanical or authenticity claims (movement caliber, water-resistance tested, box-and-papers), those slots stay photographic.
When NOT to use AI generation for this
- Water-resistance or shock-test demonstrations beyond the actual spec
- Movement caliber/serial details that must photograph the real movement
- Pre-owned condition slots (scratches are the product)
- Any render of a watch that doesn't physically exist yet
Frequently asked questions
Can AI watches fool buyers anymore?
Enthusiast communities screenshot dial typography, subdial geometry and reflection content as AI tells within seconds. Listings generated that way don't just look wrong — they read as counterfeit risk. Lock to real capture and the problem inverts: every pixel is defensible.
Do I need focus stacking?
For macro dial/bracelet detail, yes — that's the input craft. The canvas automates everything downstream of it.
What about limited-edition or pre-owned authenticity?
Condition reports, serial shots and box papers stay real photographs. Generated content supplements the presentation layer only.
Can you do lume shots?
Derived from a real long-exposure capture of the actual compound — color and coverage must match the physical dial, so the graph transforms the reference, it doesn't imagine the glow.
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.