// the find
lk-geimfari/mimesis
Mimesis is a fast Python library for generating fake data in multiple languages.
Mimesis is a Python library for generating fake data across 46 locales, aimed at developers who need test fixtures, seed data, or mock payloads. It's faster than Faker and has typed providers with good editor support. The target audience is Python developers writing tests or building data pipelines that need realistic-looking input.
Performance is the main differentiator — benchmarks show it's meaningfully faster than Faker for bulk generation, which matters when seeding a database with millions of rows. The locale coverage is genuinely wide (46 locales including Arabic variants, CJK languages, Eastern European), not just a thin English wrapper with a flag on top. Schema-based generation with relational references is a real feature: you can define interdependent schemas and get consistent foreign-key-style links across generated records. Full type annotations throughout means autocomplete actually works, which matters more than it sounds when you're hunting for `person.blood_type()` vs `person.blood_group()`.
Version 20.0.0 is announced to break schema-based generation entirely with no backward compatibility — that's a significant risk if you're building any non-trivial fixture pipeline around it today. Dropped support for built-in providers in 19.0.0 without much ceremony suggests the API surface is less stable than a library at 4800 stars should be. There's no built-in way to generate statistically realistic distributions (e.g., age skewed toward working adults, names weighted by frequency) — everything is uniform random, which produces obviously fake-looking aggregate data. Integration with ORMs like SQLAlchemy or Django ORM is manual; factory_boy integration exists but it's thin compared to what model-baker or factory_boy's own Django support provides.