Use BCILattice before you download it
This is the desktop app's real interface, running in your browser. Design a paradigm, walk the eleven-step Analysis Suite, wire a pipeline and a workflow — with the same blocks, the same steps and the same method lists the app ships.
It does not compute anything. There is no local analysis service, no GPU and no recording in a web page, so every figure and number you see is a saved result from a demo study and is marked as one. Everything structural is real; nothing measured is.
Starting BCILattice…
Click anything. Use the fullscreen button in the title bar for more room — Esc exits.
Three modules worth your five minutes
These are the three the app is actually used for, and the three the demo builds in full.
Neural Flow
Build a paradigm out of the 28 real block types — rest, motor imagery, cues, feedback, markers, loops. Edit labels, label codes and durations, nest sub-blocks inside a loop, then Compile and read the sequence back.
Add a block, set its duration, press Compile.
Analysis Suite
All eleven steps, in their two real groups. Every Source dropdown offers exactly the stages that step may consume, and every method list is the real one: the four channel-quality methods, the eight scalers, the twenty channel-selection criteria, the 71 feature families.
Open Time-Frequency and page through its seven views.
ML Suite
Pipeline, Workflow and Execute. The palette filters by provider — MNE, PyTorch, scikit-learn, XGBoost — over the same catalogue the app loads, and Compile runs the real validation rules: a pipeline needs an Input Block, a workflow needs a trainer fed Dataset, Labels and Model.
Drag an output port onto an input port, then Compile.
Subjects, files, trial counts, detected events.
Filter band, notch, reference — with the PSD before and after.
Runs, metrics, and the provenance behind each one.
Study-wide metrics, plus the Validation tab and its group statistics.
What is real here, and what is not
A demo of a research tool that fakes a measurement is worse than no demo. So the line is drawn explicitly, and drawn in the interface too.
Real
- The full navigation, in the app’s own order and sections
- The 28 Neural Flow block types, with their real colours and special codes
- The ML Suite catalogue: 5,558 blocks, grouped and filtered by provider
- All 11 Analysis steps, their stages, and their input rules
- Every method list, parameter range and default
- Compile-time validation, running the same rules the app runs
Not in the browser
- No signal is filtered, decomposed, or transformed
- No model is trained and no metric is computed
- Figures and tables are saved results, labelled where they appear
- No file access, no GPU, no local analysis service
- No account, no cloud sync, no community import
The version that computes runs on your machine
Same interface, with your recordings, your GPU, and a local analysis service that never sends your data anywhere. Free to start, on Windows, macOS and Linux.