Faker, Seed Data & Fixtures

Generic fakers

• ★ @faker-js/faker — the standard; ~60 locales; thousands of generators. Default.
• chance — alternative; smaller; older.
• casual — older still; less maintained.
• @anatine/zod-mock — generate mock data straight from Zod schemas. The pairing for any Zod-typed app.
• @ngneat/falso — alternative to faker; locales; tree-shakeable.

Schema-driven seed data

• ★ Drizzle Seed (drizzle-seed) — generates realistic seed data from your Drizzle schema; the default if you're on Drizzle.
• ★ Snaplet Seed — open source as of 2025; declarative seed graphs from Postgres schemas. Best in class.
• Prisma seeds — prisma/seed.ts + faker; conventional pattern.
• @vidalsasoon/objection-seed — for Objection.js.

Synthetic / production-shaped data

• ★ Snaplet Cloud — clones production into anonymized dev databases; the gold standard. Free tier.
• Tonic.ai — paid; enterprise; very accurate.
• Privacy Dynamics, Mostly AI — synthetic data with statistical fidelity.
• pg_dump | sed — DIY; works for tiny apps.

Fixture / factory patterns

• ★ fishery — TypeScript-first factory builder.
• factory-bot-typescript — port of Ruby factory_bot.
• @anatine/zod-mock — see above.
• rosie — older test factory.
• Mock Service Worker (MSW) + faker — common combo for API-mocked tests.

Realistic-shape generators

• fakerator, @faker-js/faker.helpers.fake — templates with placeholders.
• @stoplight/json-schema-sampler — sample data from JSON Schema.
• JSF (JSON Schema Faker) — generate from JSON Schema (or Zod via zod-to-json-schema).
• OpenAPI examples — Spectacle / prism will generate sample bodies from your OpenAPI spec.

Specialized

• @faker-js/faker.image.url + picsum.photos — random placeholder images.
• DiceBear — generate avatars from a seed.
• identicon-js — GitHub-style identicons.
• boring-avatars — beautiful generative avatars.
• @ngneat/falso finance / vehicle / commerce locales.

Generating realistic but consistent data

• ★ Seed your faker. faker.seed(123) makes runs deterministic; reproducible test failures.
• Reference IDs — when generating related rows, pass parent IDs explicitly.
• Skew toward edge cases — most fakers produce average-looking data; mix in long names, RTL strings, emoji, very large numbers.
• Localize — faker.locale = 'ja' to pressure-test non-Latin layouts.

Cleaning up between tests

• Drizzle / Prisma transactions per test — wrap each test in a transaction that rolls back.
• pg_truncate / Drizzle delete — bulk wipe in beforeEach.
• Test container per worker — Vitest pool, each gets its own DB; slowest but most isolated.
• Snaplet snapshots — restore a known seed snapshot.

Pick this if…

• Default fake data: @faker-js/faker.
• Already on Zod: @anatine/zod-mock.
• Drizzle schema seeding: Drizzle Seed.
• Realistic prod-shape dev DB: Snaplet.
• Test factories: fishery.
• Generative avatars: boring-avatars or DiceBear.