New 0.9: Workflow templates & macro mapping

Research should be intuitive.

Wire filters, decomposition and models on a canvas, inspect every intermediate result, and hand the whole pipeline to a colleague as one file. Built on the same methods as MNE and EEGLAB.

v0.9.2 · Open source, MIT · Runs locally, no account · Other platforms and the beta

Every step of the analysis, on one canvas.

Each card is a function you can open. Change one knob and every result downstream updates while you watch.

See what the filter actually did

Every card keeps its output. Click a wire and the spectrum before and after appears, with the passband drawn on it. No print statements, no re-running a script to check one step.

How execution works
The Band Pass, Notch and CAR cards, wired on the canvas

Checked against MNE, number by number

Cards that claim to reproduce a reference are run against it on public recordings. The agreement is published next to the card, so a reviewer can check it without trusting us.

Read the validation report
CardReferenceAgreement
Band PassMNE FIR filterr = 0.99
Low PassMNE FIR filterr = 0.9998
High PassMNE FIR filterr = 0.9963
Bipolar Referencemne.set_bipolar_referencer = 1.00
Common Average ReferenceMNE average referencer = 1.00
Crowther Artefact Removalartefact-removal referencemax Δ 1.9e-9

Hand the pipeline over as one file

A pipeline saves to a single file with every card and parameter. A colleague opens it and gets the same result on their machine, with no environment to rebuild.

Sharing a workflow

See it in action

The canvas, the gears and the plots, as they look on your screen.

A pipeline of seven cards on the builder canvas
A pipeline of seven cards, wired by hand
The gears of the Band Pass card
The gears tab, with every resolved value visible
The visualizer showing EEG channels
The visualizer, one trace per EEG channel

A card is just a Python function.

Decorate a function and it becomes a block with typed inputs, outputs and controls. No build system, no YAML.

my_bandpass_node.py
PYTHON
# Drop this in ~/.signal-studio/nodes/. It shows up in the palette.
from signal_studio import node, Signal
from scipy.signal import butter, filtfilt

@node(category="filter", color="#0f62fe")
def bandpass(
    x: Signal,
    low: float = 1.0,
    high: float = 40.0,
    order: int = 4,
) -> Signal:
    b, a = butter(order, [low, high], fs=x.sfreq, btype="band")
    return x.with_data(filtfilt(b, a, x.data))

Version 0.9.2

Free, local, no account and no telemetry. Signal Studio is in public beta: usable, but still moving.

All downloads

Windows

Available

x64 / .msi installer

Native build for Windows 10 and 11. The installer is not code-signed yet, so Windows shows a warning on first launch.

Download .msi

Windows 10 + · 8 GB RAM recommended

macOS

In progress

Universal / .dmg

Universal build for Apple Silicon and Intel. The code compiles; packaging and notarisation are in progress.

See the roadmap

macOS 12 + · 8 GB RAM recommended

Linux

In progress

x86_64 / AppImage

Compiles on Linux, but is not packaged or tested across distributions yet.

Packages and Docker

glibc 2.31 + · 8 GB RAM recommended

Beta

Public beta

Windows, for now

Real software you can run on real recordings today. Expect bugs, and record the version you ran next to your data.

Join the beta

Report what breaks on GitHub or Discord

Open your first recording in about ninety seconds.