# Unlock AI training data with the open-sourced Synthetic Data SDK

**URL:** <https://discuss.huggingface.co/t/unlock-ai-training-data-with-the-open-sourced-synthetic-data-sdk/138903>\
**Category:** Show and Tell\
**Created:** [February 4, 2025, 9:48am UTC](https://discuss.huggingface.co/t/unlock-ai-training-data-with-the-open-sourced-synthetic-data-sdk/138903 "2025-02-04T09:48:08Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![spintronic](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/spintronic/32/40360_2.png) [@spintronic](https://discuss.huggingface.co/u/spintronic)\
**Post date:** [February 4, 2025, 9:48am UTC](https://discuss.huggingface.co/t/unlock-ai-training-data-with-the-open-sourced-synthetic-data-sdk/138903/1 "2025-02-04T09:48:08Z")

</div>

[MOSTLY AI](https://mostly.ai/) has open-sourced its powerful Synthetic Data SDK, enabling you to create privacy-preserving, AI-generated synthetic data directly from your existing datasets - all within your secure environments.

✨ Key Features:

✅ Broad Data Support: Handle mixed data types (categorical, numerical, geospatial, text), single/multi-table datasets & time-series data.

✅ Multiple Model Types: Leverage TabularARGN (SOTA for tabular data), fine-tuned HuggingFace models, and efficient LSTM for text generation.

✅ Advanced Training Options: CPU/GPU support, differential privacy, and real-time progress monitoring.

✅ Automated Quality Assurance: Built-in fidelity & privacy metrics with detailed HTML reports for visual data analysis.

✅ Flexible Sampling: Upsample data, generate conditionally, rebalance segments, impute context-aware values, ensure fairness, and control outputs via temperature adjustments.

✅ Seamless Integration: Connect effortlessly to external databases & cloud storage with a fully permissive open-source license.

💻 Check out the SDK on GitHub: [GitHub - mostly-ai/mostlyai: Synthetic Data SDK ✨](https://github.com/mostly-ai/mostlyai)