Train and augment AI models
Create labeled examples with known physiology, events, artifacts, and timing. Balance underrepresented conditions or prototype a pipeline before real data is ready.
Synthetic biosignal data for AI & algorithm teams
Train smarter. Test harder.
Generate labeled, configurable ECG, PPG, and HRV datasets for AI/ML training, augmentation, algorithm QA, regression, and edge-case stress testing—with known ground truth and reproducible seeds.
Open registration · No patient data required · Your algorithm stays local
ECG, PPG & HRV
Arrhythmias & morphology
Noise, artifacts & sensor faults
0.01 s–24 h duration
100 Hz–1 MHz sampling
Ground-truth annotations
Deterministic variants
Local verification
From synthetic data to measurable evidence
- AI training & augmentation
- Edge-case test sets
- Reproducible regression QA
- Local scoring
Why Synsigra
Useful data, not just plausible waveforms
Build training sets, controlled benchmarks, and hostile test cases around the question your team needs to answer—without waiting for the right real-world recording to appear.
Test the cases that matter
Turn rare rhythms, difficult morphology, low signal quality, sensor failures, and deliberate artifact combinations into repeatable tests.
Know the correct answer
Use exact annotations and packaged ground truth to train supervised models, score detector output, and understand precisely where performance changes.
Control the full experiment
Set duration, sampling, physiology, amplitude and frequency modulation, population variation, artifacts, targets, and deterministic randomization.
Scale from seconds to 24 hours
Run a quick smoke test or generate long, time-varying workloads with changing rate, HRV structure, respiration, activity, PPG timing, and perfusion.
Reproduce, compare, automate
Pin every package to its seed and generator version, compare runs with machine-readable reports, and automate generation through the API.
Curated packs
Start fast without giving up control
Pick a proven starting point for a common job, generate it as-is, or customize duration, sampling, physiology, modulation, artifacts, and targets when your experiment needs more.
Popular
v1
R-Peak Stress
Smoke and stress verification for R-peak detection under clean and artifact-heavy conditions.
- Scoreable: R-peak, signal quality
- Best for: detector regression
Benchmark
v1
HRV Benchmark
Evaluate HRV pipelines using synthetic scenarios with known rhythm and exclusion behavior.
- Scoreable: HRV, quality
- Best for: metrics QA
Rhythm coverage
v1
ECG Rhythm & Arrhythmia
AF, flutter, SVT/VT, PAC/PVC, pauses, AV block, bundle-branch cases, and transition annotations.
- Scoreable: rhythm and event outputs
- Best for: arrhythmia scenario coverage
Classification
v1
ECG Morphology & Beats
P-QRS-T morphology, fiducials, beat labels, conduction morphology, QT/ST-T behavior, and phenotype stress.
- Scoreable: beat classes, morphology metrics
- Best for: ECG QA and population variation
Quality
v1
Signal Quality
Artifact intervals, quality masks, and visibility into scoreable versus reference-only pack outputs.
- Scoreable: quality outputs
- Best for: acquisition QA
Wearable
v1
Wearable QA
ECG/PPG alignment, PPG peaks, low perfusion, motion/sensor artifacts, long-duration cases, and multi-signal workloads.
- Scoreable: PPG, ECG/PPG alignment, HRV
- Best for: wearable system QA
How it works
From a question to usable data in four steps
Generate once for training, generate repeatedly for controlled variation, or keep a fixed package as a regression benchmark. The same workflow supports both data creation and algorithm verification.
Example verifier flow
python -m pip install synsigra-wheel.whl
synsigra-verify . detections/ verification-results/ \
--profile regression --forceChoose the job
Start with a curated pack for AI data, detector QA, HRV, morphology, signal quality, PPG, or wearable testing.
Configure and generate
Use safe defaults or tune duration, sampling, physiology, modulation, artifacts, targets, and randomization.
Use the data locally
Train a model, augment a dataset, run a detector, or feed the signals into your existing pipeline on your own machine.
Measure and reproduce
Score supported outputs, archive machine-readable evidence, and rebuild the exact package whenever you need it.
Built for focused work
Simple when you start, precise when you need it
Follow a guided pack flow for common goals, then open the full scenario editor only when you need detailed control. Every choice stays tied to the generated package and its evidence.
Guided startChoose a goal and a curated pack
Full controlCustomize physiology, timing, variation, and artifacts
Clean handoffDownload one self-contained package with data and verification tools
Developers
Generate through the UI or automate through the API
Create the same versioned, deterministic packages from scripts, agent workflows, notebooks, CI jobs, or your internal data pipeline.
Example create-job request
curl -X POST /syn_sig_ra/v1/jobs \
-H "Authorization: Bearer <api-key>" \
-H "Content-Type: application/json" \
-d '{
"project_id": "default",
"pack_id": "r_peak_stress_v1"
}'Self-describing packages
Manifests, fingerprints, target metadata, and pack summaries make every downstream input explicit.
Automation without ambiguity
Pin pack and generator versions, poll job state, download one artifact, and reproduce the same dataset later.
Discoverable API
Use the machine-readable OpenAPI contract to discover authentication, packs, jobs, downloads, and account workflows.
Private by design
Your model and algorithm stay yours
Synsigra generates the signals and ground truth. Your training, detector execution, model weights, and proprietary code remain in your own environment.
No patient data required
Create training and test inputs from declared synthetic parameters without uploading personal health information.
Local execution
Train models and run proprietary detectors privately; downloaded verification tools also run on your machine.
Traceable inputs
Package manifests, generator provenance, fingerprints, summaries, and reports make test inputs auditable.
Clear validation boundary
Synthetic data accelerates engineering and complements real-world data; it does not replace clinical or regulatory validation.
Resources
Everything needed to move from first pack to automation
Explore capabilities in the product, inspect every pack before generation, and use the verifier and API contracts for repeatable workflows.
Guided product
Choose a goal, compare packs, customize when needed, and generate from a clear step-by-step flow.
Explore the productPack contracts
See intended use, data coverage, targets, scoreability, and reference-only outputs before generating.
Browse live packsLocal verification
Each downloadable kit includes detector templates, exact-version tools, recipes, and machine-readable summaries.
See the workflowDeveloper API
Discover authentication, capabilities, packs, scenarios, jobs, artifacts, and account operations.
Open the API contractReady to build better data?
Give your model more than the easy cases
Generate a labeled dataset in minutes, then scale from clean baselines to long-duration, modulated, artifact-heavy scenarios. Start curated and customize only when you need to.