Problem
What needed solving
A practical metadata workflow that understood marketplace-specific fields, keyword limits, export formats, GenAI disclosure, and common compliance risks.
Metadata Tool
MetaGen
A bring-your-own-key workbench for generating reviewable stock and SEO metadata from images.
Product story
Why it exists.
MetaGen supports stock contributors and image SEO workflows by turning visual analysis into editable titles, descriptions, ordered keywords, warnings, and export-ready files.
Origin: Replace generic AI tagging with a review layer designed around the actual fields and constraints contributors work with.
Built to shorten repetitive metadata work without removing contributor review, platform rules, or data ownership.
Built
What we built
An accountless React workbench with multi-provider image analysis, Adobe and Shutterstock CSV exports, iStock/Getty drafts, SEO JSON, and deterministic compliance checks.
Outcome
What exists now
A tested metadata product with a polished responsive workbench, four output profiles, custom profiles, and 23 passing behavior contracts.
Verification
Proof you can inspect.
Verified against the active MetaGen repository, local workbench, test suite, and TypeScript build on July 15, 2026.
- 01 The workbench supports Adobe Stock, Shutterstock, iStock/Getty draft, General SEO, and custom output profiles.
- 02 Provider adapters cover Gemini, OpenAI, OpenRouter-compatible APIs, and Anthropic while keeping BYOK settings under user control.
- 03 The project passes 23 unit tests covering UI contracts, API errors, compliance rules, exporters, custom profiles, and provider parsing.