Build analysis pipelines by wiring nodes, turning industrial acoustics and NVH analysis from writing code and tuning black boxes into a workflow you can read at a glance.
Single-machine workstation or on-premise server, the core is identical: wire nodes into pipelines, let AI orchestrate, run professional acoustic algorithms.
Drag and connect to form a pipeline; complex analysis becomes one readable diagram.
State your intent in one sentence; AI builds the pipeline and runs the tests.
Mainstream psychoacoustic and NVH metrics, with physical quantities preserved end to end.
Drag, connect, run, and get results.
Drop in nodes and wire them into a DAG; spectra, waveforms, and metrics land right on the nodes, what you see is what you get.
Cached by input fingerprint and reused in seconds; stop a run at any time with one click.
Every change can be undone and redone; jump back to any step from the history dropdown.
Save tuned parameters as named presets, apply them in one click, share by code or export to migrate.
Fork from any snapshot to iterate independently; switching, rolling back, and comparing stay isolated.
Manage workflows with three views and nested folders; switch tabs without losing state.
Channel semantics, physical quantities, and calibration are fixed on ingest and carried through every stage.
Upload WAV/NPZ/CSV/TDMS directly, with recursive extraction from folders.
Compose single-channel audio into multi-channel with three pairing modes.
Store name, unit, physical quantity, and calibration per channel; attach a template to apply automatically.
Declare sound pressure / acceleration and sensitivity per channel; the dB reference defaults to the standard value.
Extract or split single channels and process each independently without averaging.
Vibration signals are not re-normalized, so metrics match the acquisition instrument.
Preprocessing · acoustic metrics · frequency-domain analysis · feature engineering, covering the backbone of NVH analysis. Every analysis node ships with industry presets, ready to run in one click. See the full matrix in Nodes.
Presets map to industry scenarios or public standards (such as “Bearing · General” or “One-third Octave”) — beginners pick a preset, tweak one or two options, and run; professionals can fine-tune every stage, and the parameters can be brought into the AutoML search.
Search parameters automatically, distill into an explainable formula, and score in one click.
Pick data and tunable nodes in four steps; start by forking automatically.
Optuna Bayesian search with early stopping for poor trials.
Classifier, spec limits, similarity — deciding OK/NG.
Live learning curves make parameter importance plain to see.
Fit the black box into a LaTeX discriminant formula, explainable.
Generate a pipeline from the formula in one click; wire it up to score.
Overfitting at a glance — every trial logs both training and validation scores, and a gap beyond the threshold is automatically flagged as an overfitting warning; in the diagnostics view, clicking any sample opens an audio player and spectrum chart so you can see “why this one was misclassified.”
Viewers preview instantly; drag them into a dashboard to build an interactive report.
Write your own nodes; publish, subscribe, and call them — you set the boundaries.
Describe your intent and AI will build pipelines, run tests, and write nodes.
Authorize a client once via OAuth; permissions are isolated by scope.
Add nodes, wire connections, and change parameters in one sentence — and even write new nodes.
Watch what AI is doing in real time on the activity panel; the process is fully transparent.
When AI is uncertain, it asks first and waits for your choice before continuing, never deciding on its own; on the desktop workstation, the local AI client connects directly without authorization — just install the desktop plugin to use it.
Nodes, AutoML, AI, and visualization are shared by both editions. The only differences are deployment form and collaboration scale — see them item by item below.
One server, the whole team analyzing online at once.
Organization + user groups + per-feature authorization, with data isolated by owner.
Workflows / data sources / templates accumulate and are reused across the team; reproduce a methodology in one click.
Centralized intranet deployment; open a browser and go, no installation.
Both editions share the same workflow format, and workflows migrate freely between workstation and server via .tinpack files.
Each release distills one capability, making analysis faster and more stable.
Heavy computation runs on a shared GPU sidecar for acceleration; if unavailable it falls back to CPU automatically with identical numerics.
Outputs features such as dominant scale and scale entropy; switch coarse to fine with one slider.
The SDK keeps pushing data and reading results in real time, well suited to online monitoring.
Processes stay resident and libraries load only once; high-frequency calls reuse the warm process.
A cache hit shows the time saved, with reuse persisting across runs.
Fork from any snapshot to iterate independently, with rollback and comparison fully laid out.
All of the above capabilities have shipped in official releases. Finer version-by-version changes can be reviewed in the in-product Changelog.
Request a trial and experience the complete path from raw signal to an explainable verdict. We will recommend the right deployment form based on your industry and scenario.