← Back to workNext Up:Retail Intelligence Dashboard

Product Image Platform

Visit foxs.com ↗
GeminiOpenAIVision ModelsParallel Processing

Team

5-person daily workflow

Cost

$100K+ annual cost saved

Time Saved

2000+ hours a year

Scale

5K+ annual SKUs

AI-powered product imagery workflow used daily by a 5-person e-commerce team. Transforms vendor product photos into e-commerce-ready assets for foxs.com, replacing outsourced model photography and eliminating ~$100K in annual production cost.

Problem

  • Product imagery relied on expensive, slow-to-schedule models.
  • Coordinating shoots pulled the e-commerce team into manual production work.
  • Each new arrival waited on a model shoot before it could go live.

Solution

  • AI workflow that converts vendor photos into on-brand product assets.
  • Front, back, on-model, and standalone variants generated in parallel.
  • Roughly 10x faster and ~100x cheaper than outsourced model shoots.

System Overview

01

Batch Uploads

Product images come in as iOS photo uploads or directly from Fox's centralized image database, then get grouped by SKU for production.

02

Parallel Processing

Gemini processes batches in parallel, generating front, back, on-model, and standalone product assets from the selected source imagery.

03

Human In The Loop

The team can generate, review, edit, and refine each image with prompting before approving assets for commerce use.

04

Push To Shopify

Approved assets are pushed into Shopify with the right product context so they can move into the live storefront workflow.

Challenges

  • Cohesive, consistent outputs: a structured prompt library and a precategorization step that picks the right prompt per item.
  • Model drift: a model generator and library to manage and iterate on AI models as they shift.
  • Human refinement: not perfect on the first pass — a quick interface to retry and apply prompted edits to specific outputs.