SynthCity
github.com
- Free plan
- Linux
What SynthCity is
SynthCity is a free Python library for generating and evaluating synthetic data, available for Linux and self-hosted use. Its generators cover tabular data, time series, survival analysis, and images, with listed approaches including GANs, VAEs, normalizing flows, Bayesian networks, random forests, and LLM-based generators. A plugin architecture is designed to support extensions. Evaluation metrics address data correctness, quality, and privacy, while the Benchmarks API can assess generators using selected metrics and repeated runs. Privacy-oriented methods include DECAF, DP-GAN, AdsGAN, PATEGAN, and PrivBayes; DECAF also supports inference-time debiasing based on user-defined fairness requirements. The README demonstrates use with scikit-learn and lifelines datasets. The library can be installed from PyPI or source, and its utilities serialize and load models from files or buffers. API access is available. The project identifies its license as Apache-2.0, tracks public bugs through GitHub Issues, and welcomes pull requests. SynthCity does not handle missing data, so it must be imputed before use; HyperImpute is suggested for that task. Its listed image generator architectures are not state of the art.
Who it suits
SynthCity suits Python users who need to generate and evaluate synthetic datasets across several data modalities. Its fairness and privacy methods may be relevant to teams with those evaluation needs.
Good to know, and what to watch
- Generators cover tabular, time-series, survival, and image data
- Evaluation metrics address correctness, quality, and privacy
- Benchmarking supports selected metrics and repeated runs
- Plugin architecture is designed for extensions
- Apache-2.0 license; install from PyPI or source
- Missing data must be imputed before use
- Listed image generator architectures are not state of the art
Plans
Every plan and priceSynthCity
Free plan
The verdict
SynthCity offers a broad set of synthetic-data generators and evaluation tools for self-hosted Python workflows. Plan to impute missing data first, and consider the stated limitation of its image generator architectures.
What the maker publishes
- Benchmarking
- The Benchmarks API can evaluate generators using selected metrics and repeated runs. source
- Data modalities
- The repository describes generators for tabular data, time series, survival analysis, and images. source
- Evaluation
- It includes evaluation metrics for data correctness, quality, and privacy. source
- Extensible architecture
- The library has a plugin architecture designed to be extended. source
- Fairness and privacy
- Privacy-focused methods include DECAF, DP-GAN, AdsGAN, PATEGAN, and PrivBayes, and DECAF supports inference-time debiasing using user-defined fairness requirements. source
- Generator types
- The listed models include GANs, VAEs, normalizing flows, Bayesian networks, random forests, and LLM-based generators. source
- Image model limit
- The README says its image generator architectures are not state of the art and suggests extending the image generator/discriminator architecture method for other architectures. source
- Installation
- The library can be installed with pip from PyPI or installed from source. source
- Integration
- The README demonstrates using SynthCity with scikit-learn datasets and lifelines survival datasets. source
- License
- The repository identifies its license as Apache-2.0. source
- Missing data limit
- SynthCity does not handle missing data, which must be imputed first; the README points to HyperImpute for this purpose. source
- Model serialization
- Models can be serialized and loaded from buffers or files using the library's serialization utilities. source
- Purpose
- SynthCity is a library for generating and evaluating synthetic tabular data. source
- Support and contributions
- The project tracks public bugs through GitHub Issues and welcomes pull requests. source
Checked against the maker's pages on 3 Oct 2026.