# microsoft/TimeCraft

Official code for TimeCraft: A Time Series Generation Framework for Real-World Applications

Repository: https://github.com/microsoft/TimeCraft
Canonical: https://ross.abutalabs.com/products/timecraft
Language: Python
License: MIT
License Family: permissive
Topics: generative-ai, machine-learning, timeseries-analysis
Last push: 2026-08-07T06:05:08+00:00

## Health v2 (maintenance only)
Score: 64/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 96, release rhythm 35, longevity 43
- inputs: {"age_days": 604, "days_push": 26, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1085, forks 65 (observed 2026-08-28T04:03:31.674080+00:00)

## What it is
TimeCraft is a diffusion model-based framework for generating high-quality synthetic time series data across domains. It uses learned semantic prototypes and a Prototype Assignment Module to adapt to new domains with few-shot examples, enabling controllable, privacy-preserving generation.

## Use cases
- generate synthetic time series data
- augment scarce time series datasets
- create privacy-preserving synthetic sensor data
- simulate time series for forecasting model training
- generate domain-specific time series with few-shot examples
- produce controllable synthetic trends and seasonality

## When to choose
- you need synthetic time series for domains with limited real data
- privacy constraints prevent sharing real time series
- you want controllable generation of trends or seasonality
- you need cross-domain generalization without retraining from scratch

## When to avoid
- you need simple classical time series forecasting rather than generation
- you require production-hardened tooling rather than research code
- your data is not temporal in nature
- you lack GPU resources for diffusion model training or inference

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, data-generation, deep-learning
- domain: time-series, machine-learning, data-science
- platform: python, cross-platform
- tags: diffusion-models, synthetic-data, time-series-generation, generative-ai, research-code

## Member repositories
- microsoft/TimeCraft (main) score 64

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:31.674080+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:49:52.125891+00:00, confidence not recorded.
  - readme: https://github.com/microsoft/TimeCraft (fetched 2026-08-28T04:03:31.674080+00:00, sha 6277c44b2cd9)
- Data as of 2026-08-30T08:39:29.467469+00:00.