I'm building SeedBase. The itch: every project I join, the dev/test database is either three hand-typed rows that test nothing, or a copy of production with real customer data on everyone's laptop. And the usual fake-data generators give you "valid" but obviously synthetic data: every product is "Premium Widget 1", every bio is lorem ipsum, a blog title has nothing to do with its body.

So SeedBase reads your schema (a SQL dump, Django models.py, or a Prisma schema) and generates synthetic data where every foreign key resolves and the data actually looks real: names match emails, order totals match their line items, realistic distributions so pagination and N+1 bugs surface. Multi-database, deterministic per seed, and there's a login-free sandbox where you paste a schema and see the output in your browser.

The part I sank the most time into was making the fake data not look fake, and parsing messy real-world schemas. I tested it against my old 20-app Django project with 226 tables. Still early, would love feedback from people who fight with test data.

You can paste a schema and see the output in the browser, no signup needed. seedba.se/playground