Lyrebird · Realistic synthetic test databases, generated fully offline
SUBSETRAPLATFORM

Lyrebird · Offline Synthetic Test Data

Synthetic test data.
Generated 100% offline.

Connect to your database once to create a compact mathematical recipe. Disconnect completely, then generate millions of realistic, foreign-key-consistent test records 100% offline without touching production again.

Request a demo Explore Ark Platform
  • 100% offline & local execution
  • Preserves foreign-key graphs
  • PostgreSQL, MySQL, SQL Server & Oracle
  • Desktop & CLI

01 Connect once. Save the recipe.

Source database One-time read-only
Profile file Compact mathematical recipe

Statistical distributions and foreign-key graph metadata.

Source database is now disconnected. Physical air-gap.

02 Generate from recipe. Repeat offline anytime.

Saved profile
Lyrebird
Synthetic data

100% offline · Zero connection to production required

Profile once. Reuse the recipe for as many test environments as you need.

Works with the databases
you already use.

Core Capabilities

Realistic relational testing.
Zero production data exposure.

Engineered for engineering, data, and security teams who need high-fidelity test databases without the legal, compliance, and infrastructure risks of cloning production.

01 / CAPABILITY

Relational Integrity

Parent rows are generated before their children, and every child references a valid parent. Composite keys and foreign-key constraints hold across the whole schema.

02 / CAPABILITY

Fully Offline Execution

The local model and the generator run on your laptop, CI runner or private servers. No cloud service or external API is involved.

03 / CAPABILITY

Profile Privacy Audit

Before a profile leaves your team, Lyrebird lists what it keeps as written and scans it for personal data and credentials.

04 / CAPABILITY

Data Sized to the Job

Scale every table by a factor, from a 1% copy to a 1,000x load test, or set total and per-table row counts. Table ratios and foreign-key degrees are preserved.

05 / CAPABILITY

Cross-Engine Generation

Profile PostgreSQL, MySQL, SQL Server or Oracle, then generate for the same engine or another one, for example an Oracle profile into PostgreSQL.

06 / CAPABILITY

Reproducible, Ready-to-Load Output

The same profile, settings and seed give the same data. Output is SQL files with a manifest, native bulk-loader files for SQL Server and Oracle, or a database loaded straight into Docker.

A workflow built for enterprise security

Connect once.
Generate independently.

A compact profile file is all Lyrebird needs from the source database. Every test dataset after that is generated from it.

  1. 01 Connect & Profile

    Connect once & profile

    Connect to the source database read-only. Lyrebird analyzes schema, relationships, and value distributions, saving them in a compact profile file.

  2. 02 Disconnect

    Disconnect & air-gap

    Disconnect from the source database completely. Lyrebird needs no further access to your live infrastructure or sensitive production rows.

  3. 03 Generate

    Generate & test offline

    Load the saved profile and specify your target row counts. Generate synthetic relational databases anytime from the desktop app or CI pipeline.

bash — lyrebird automated pipeline

# Step 1: Profile the source database (run once, read-only)

$ LYREBIRD_DSN="$PROD_READONLY_DSN" lyrebird profile --output app.lyre


# Optional: check the profile for personal data before sharing it

$ lyrebird audit-profile --input app.lyre


# Step 2: Generate a 10% copy offline and load it into a Docker database

$ lyrebird generate --profile app.lyre --scale 0.1 --out ./testdb --docker

# ./testdb: 01_schema.sql, 02_data.sql, 03_post_data.sql, manifest.json

Community & Enterprise

Evaluate with confidence.
Expand when you’re ready.

Start with Community to evaluate Lyrebird against your own databases. Move to Enterprise when you need larger volumes and pipeline automation.

Free Evaluation

Community Edition

For development teams evaluating data realism, relational consistency, and local workflows on developer workstations.

  • Local database profiling and generation
  • Up to 5,000 synthetic rows per table
  • Native desktop app for Linux, macOS & Windows
  • Free profiling, inspection and JSON export
  • 100% offline and air-gapped execution
Request evaluation demo
Scale & Automation

Enterprise Edition

For organizations integrating high-volume synthetic test data into automated CI/CD pipelines, staging clusters, and performance testing suites.

  • No 5,000-row limit per table
  • Headless CLI for scripted pipelines and CI/CD runners
  • Override files for row counts, column and conditional distributions, and seeds
  • Offline enterprise license validation
  • Direct email contact with the engineers who build Lyrebird
Request enterprise demo

Product FAQ

Questions from your engineering team.

Everything you need to know about evaluating and deploying Lyrebird.

Can our team evaluate Lyrebird before purchasing?

Yes. Community is free and lets your team evaluate local profiling and desktop generation with your own database. Generation is limited to 5,000 rows per ordinary table. Enterprise adds larger-scale generation and headless CLI automation.

Does generation need a live database connection?

No. You connect to the source database once to create a compact mathematical recipe (profile). After that, Lyrebird generates from the saved file without reconnecting. Profiling and generation both run inside your own environment, with no cloud service involved.

How does Lyrebird preserve multi-table foreign keys?

Lyrebird analyzes database metadata and builds a relational dependency DAG (directed acyclic graph). Parent entities (e.g. users, companies) are generated first, and child entities (orders, transactions) bind to valid parent keys, so no child row is left without a parent.

Can we integrate Lyrebird into automated CI/CD pipelines?

Yes. The Enterprise edition includes a headless CLI binary for Linux, macOS, and Windows. Given the same profile and the same row settings, it produces the same data on every run, so CI test databases are reproducible. It can also load the result into a Docker database and verify row counts.

Plan your enterprise evaluation

Let’s discuss
your test data requirements.

Share your database stack, target scale and deployment requirements.
We’ll help your team assess the right fit.

Request a demo info@subsetra.com