OpenGalea is what our team built at MIT Reality Hack 2025: an affordable, open-source path from EEG to mixed reality on Meta Quest 3, designed so more than one person in the same physical room can share anchors, see the same virtual layer, and collaborate using brain-derived signals in real time. We were not trying to clone a closed lab rig; we were trying to make “neuroadaptive MR” something a small team could actually assemble, run, and extend after a weekend — then prove it in a multiplayer demo.
Why we built it
High-end neural interfaces and MR headsets have both become dramatically more capable, but the overlap between them is still thin in open hardware and open software. Commercial neuroimaging rigs are expensive and opaque; Quest-class passthrough MR is everywhere among developers, yet few projects close the loop from scalp to shared scene with clear documentation. We wanted a stack that could be reproduced: OpenBCI Cyton, a modified Ultracortex IV-inspired headset form factor mated to Quest 3 comfort and balance, Unity on the Meta XR SDK, and a Python side for acquisition and ML — with colocation so the experience is inherently social, not a solitary EEG minigame.
What it does
A user wears the EEG cap and Quest 3. Raw data is , features are computed, and classifications or band-power summaries are pushed into Unity over the network. In our showcase build, multiple users in one space cooperate in MR: each person's signal is visualized (for example as pillars tied to alpha dynamics), and the group coordinates timing so a shared character clears obstacles — a legible metaphor for “we are jointly controlling something with our state, not just our hands.” The same plumbing supports simpler single-user brain-controlled object manipulation; the hack emphasis was on colocated multiplayer MR + BCI.
Hardware
We used an OpenBCI Cyton board with an eight-channel montage, routed through a headset derived from Ultracortex IV ideas but redesigned for cost, printability, and integration with Quest 3 straps and counterweights. Getting a stable mechanical interface between a consumer HMD and a wet-electrode rig is most of the battle: comfort, slip, cable strain, and session length all feed directly into whether the demo is honest or theatrical.
Software and integration
The MR client is Unity 2022.3 with the Meta XR SDK: passthrough, colocation, shared spatial anchors, and multiplayer. Each participant runs OpenBCI tooling on a laptop; streams reach Unity via UDP (and we supported both direct alpha-style features from OpenBCI prefabs and a path through a small ML server). The point was not to hide the biophysics behind a black box — it was to show a reproducible bridge from lab streaming to game loop.
Machine learning pipeline
We collected labeled windows for two mental states, attention versus relaxation, over hundreds of thousands of samples, then trained a Random Forest on raw channels plus alpha and beta band power. A Python service pulls live data over LSL, scores one-second windows, and emits JSON predictions over UDP so Unity can react every second — slow by gaming standards, fast enough to read as intentional control for the hack narrative and for co-present spectators to understand cause and effect.
What was hard
Aligning clocks and perceived across OpenBCI, laptops, and multiple headsets; keeping classifiers honest on mobile EEG; iterating the printed headset until sessions could run long enough for a judge walkthrough; and making multiplayer MR feel fair when each brain is a different instrument. Those problems are the real product, not the shader.
Recognition
We received a Meta prize that included five Quest 3 headsets for the team — a huge accelerant for continued work. Exact hackathon labels and submission metadata live on Devpost.
Team
Team Syncer: Alif Jakir, Tsing Liu, Soo Bin Park, Yechan Ian Seo, Syed Hussain Ather — with hardware engineering support from Vankley Yi (credited on Devpost). Roles split across Unity/XR, ML and real-time inference, hardware fabrication, art direction, and glue code; the Devpost page preserves each member's self-description.
What I worked on
I pushed the project toward a coherent thesis (affordable open neuro-MR + colocation), produced design renders and hardware iterations, helped prototype and test the rig, wrote and organized documentation, and stitched subsystems together when the clock ran down — alongside teammates who owned large chunks of Unity integration, signal processing, and fabrication.
Links and media
- Devpost — OpenGalea (screenshots, full submission text)
- github.com/Caerii/OpenGalea — hardware + design
- github.com/Caerii/OpenGaleaSoftware — Unity, ML, and BCI integration (stable fork)
- Demo video · Sizzle reel
MIT Reality Hack 2023 — LifeInBetweenXR · MIT Reality Hack 2026 — Personoid / De-Escalate