Windows
Native x64 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 or later, 8 GB RAM recommendedSignal Studio / Download
Free, open and local. Runs entirely on your machine, with no account and no telemetry. Version 0.9.
Native x64 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 or later, 8 GB RAM recommendedNative build for Apple Silicon (M1 and later). Not notarised yet: the first time, right-click the app and choose Open.
Download .dmg macOS 13 or later on Apple Silicon, 8 GB RAM recommendedPackage for Ubuntu 22.04+ and Debian 12 (installs to /opt). A portable .tar.gz is listed in all files below.
Download .deb glibc 2.35 or later, WebKitGTK, 8 GB RAM recommendedReading the release feed…
Compare the SHA-256 of the file with the one in the table above before installing, especially on a shared machine.
shasum -a 256 <file>certutil -hashfile <file> SHA256Planned. They all expect a signed installer, and code signing is not done yet. Why, and the Docker image.
winget install NeuraCrypt.SignalStudioplannedbrew install --cask signal-studioplannedchoco install signal-studioplannedSignal Studio right after the first launch, and the template picker.
Run the installer on Windows, drag to Applications on macOS. Signal Studio opens on the Welcome view.
ERP CORE, sleep staging, resting-state qEEG, or a blank canvas. Click Use template.
Drag a recording onto the canvas. The format, channel types and montage are detected for you.
Press Run. Every card executes and every view updates. Open any card to check its output.
| Part | Minimum | Recommended |
|---|---|---|
| OS | macOS 12, Windows 10, Linux with glibc 2.31 | macOS 14, Windows 11, Ubuntu 24.04 |
| CPU | 4-core x86_64 or ARM64 | 8-core Apple Silicon, Ryzen or Core i7 |
| RAM | 4 GB free | 16 GB, 32 GB for multi-hour high-density recordings |
| GPU | Integrated, with Metal, Vulkan or DX12 | Discrete, for 256-channel real-time scrubbing |
| Disk | 500 MB for the install and cache | NVMe SSD for out-of-core recordings |
Yes, really. MIT licensed, with no tiers, no commercial version and no paywall. The project is funded by donations and grants. If it is useful to you and you have the budget, sponsor a contributor, but you will never be required to.
No. Signal Studio runs entirely on your machine. There is no telemetry and no analytics. It only talks to the Marketplace when you fetch or publish something yourself, which makes it fine for clinical environments and air-gapped lab machines.
Yes, that is the point. Signal Studio imports and exports MNE Raw, Epochs and Evoked objects, and reads and writes EEGLAB .set files. Many built-in cards are thin wrappers around MNE-Python.
For most research workflows it is comparable or better. Where commercial tools win is cleared clinical reporting: Signal Studio is a research tool, not a medical device. The Software page has the comparison.
Yes. App Mode hides the graph and shows only the parameters. Pick a template, drop in your file and press Run.
Every workflow has a cite-as block that credits Signal Studio and the workflow's author, with versions pinned. To cite Signal Studio itself, see the Cite page in the footer.