Metasyn

Metasyn generates synthetic tabular data from sensitive datasets, with the aim of making those datasets more accessible, reproducible, and reusable. Users fit a MetaFrame to a dataframe, save the model and settings as auditable JSON, then generate new rows. It handles varied data types, missing values, categories, unique keys, and structured strings such as postcodes and identifiers. Metasyn supports pandas and Polars dataframes and uses Faker to produce localized names, emails, phone numbers, and other data. By default, its model omits multivariate information to reduce certain disclosure risks. A disclosure-control plugin is available, but generated data still needs human review and is not guaranteed to be free of sensitive information. Users can add extensions through a plugin system and export results to CSV, Parquet, Excel, and other formats. The free, MIT-licensed Python package is self-hosted on Linux, macOS, and Windows, with installation through pip or GitHub.

Metasyn offers a flexible way to generate synthetic tabular data with dataframe support, configurable extensions, and several export formats. Its output is not guaranteed to be free of sensitive information, so human review remains necessary.

Metasyn is intended for public organizations, research groups, and individual researchers who own sensitive datasets. It suits users who can work with Python dataframes and manage a self-hosted tool.

Metasyn is a free, self-hosted Python package for generating synthetic rows from tabular data. It is best suited to researchers, public organizations, and individual researchers who control sensitive datasets and need a repeatable workflow. Choose it for auditable, extensible generation in Python; look elsewhere if you need masking, a managed service, or privacy guarantees without manual assessment.

Overview

Metasyn fits a MetaFrame to a dataframe, saves the model and its settings as auditable JSON, and then synthesizes new rows. That sequence gives teams a configuration they can inspect, preserve, and reuse, which is useful for reproducible research and controlled data preparation rather than one-off sample files.

The package works with missing values, categorical columns, unique key fields, and structured strings such as postcodes or identifiers. Its default model excludes multivariate information to reduce identity, attribute, and group-disclosure risks. A disclosure-control plugin adds another privacy-oriented option, but generated data is not guaranteed to be free of sensitive information. Human review and a separate disclosure assessment remain necessary.

Key features

Polars and pandas dataframe support lets a team keep its existing Python data stack instead of converting everything to a new format. Faker supplies localized names, email addresses, phone numbers, and other values, helping synthetic records retain realistic formats across locales.

A plugin system allows users or organizations to add extensions for their own requirements. That is a meaningful advantage for institutions with specialized schemas, but it also means implementation, maintenance, and validation are the user's responsibility. Generated dataframes can be exported to CSV, Parquet, Excel, and other formats, so the output can move into common analysis and exchange workflows.

Installation is available through pip from PyPI or directly from GitHub for the development version. Versioned Docker containers and a command-line interface support repeatable deployment and programmatic use, while API access can connect generation to broader tooling. Support is centered on GitHub issues and contact with Raoul Schram rather than a managed service desk.

Pricing

Metasyn costs nothing. The single Metasyn plan is an open-source Python package under the MIT License, permitting free use, modification, distribution, and sublicensing subject to the license conditions. There is no free trial because the software is already free, and no seat, quota, renewal, or export cap is stated.

This model suits teams able to run and support software themselves. The trade-off is that the price does not include hosted infrastructure, privacy review, or operational assistance; those obligations stay with the user.

Platforms

Metasyn runs on Linux, macOS, and Windows, with self-hosted deployment. Its Python package, Docker containers, command-line interface, and API access make it suitable for local workstations, internal servers, and automated pipelines, provided the organization manages its own environment.

Who it's for

Public organizations, research groups, and individual researchers who own sensitive tabular datasets are the clearest fit. They get reproducible model settings, support for common dataframe libraries, and export formats that can feed downstream analysis or sharing. It is less suitable for teams seeking data masking—the package does not provide it—or for users who want a hosted interface that handles privacy decisions and infrastructure.

Pros and cons

Pros

• Auditable generation: Saving model and settings as JSON makes runs easier to inspect and reproduce.
• Broad tabular handling: Missing values, categories, unique keys, and structured strings are covered in one workflow.
• Python flexibility: pandas, Polars, Faker localization, plugins, API access, and CLI deployment support customized pipelines.
• Open-source economics: The MIT-licensed package has no purchase price and can be modified or redistributed under its conditions.

Cons

• Privacy is not automatic: Even with the default approach and disclosure-control plugin, output requires human review.
• No masking or subsetting: Teams needing those operations must add other tools to their workflow.
• Self-hosting burden: Users provide deployment, maintenance, and technical support resources themselves.

Alternatives

For a broader directory of options, see Test Data Management Software.

Windocks is a better fit when database containers and a hosted or API-capable deployment model matter more than a Python dataframe workflow; its free tier limits database image size to 1 GB and simultaneous containers to one.

Choose Data Maestro if a 30-day full trial and a graphical test-data workflow are priorities; its free tier then limits exports or database pushes to 100 rows and saves two database connections.

FakerForge suits teams wanting web, API, and multi-platform access with metered free allowances, including 100 rows per table and 500 API requests on its free plan.

Pick Mockaroo for browser-oriented generation with 1,000 rows per file and 200 API requests per day on its free tier.

Synthetic Data Vault is an alternative for users who want a community edition with five data types, five basic constraints, nine models, and CSV/Excel support.

Tonic Fabricate fits teams seeking a cloud-backed paid product with a free tier and monthly credits.

Databucket is another free, open-source self-hosted option.

Faker.js is the simpler choice when you only need a MIT-licensed web data-generation library rather than dataframe modeling and synthesis.

Verdict

Metasyn is a strong choice for technically capable researchers and public-sector teams that need reproducible synthetic tabular data in a self-hosted Python environment. Its combination of dataframe coverage, auditable JSON settings, extensibility, and practical exports is the main reason to choose it. Look elsewhere when you need masking, managed operations, or a guarantee that generated records contain no sensitive information.