Most questions are answered below. The rest go to a human.

Almost always one of three things: microphone permission, wrong device selected, or a hardware mute on a headset.

• macOS:System Settings → Privacy & Security → Microphone → enable NeuroBridgeEDU. Restart the app afterwards, the permission handshake doesn't apply mid-session.
• Windows:Settings → Privacy & Security → Microphone → enable NeuroBridgeEDU.
• On the Home screen, open the device dropdown and pick your microphone by name — not "Default".

Expected speed is roughly 0.3×–1× recording duration depending on your hardware and the chosen Whisper model. If yours is much slower:

1. Switch to a smaller Whisper model in Settings → Transcription. Try small or base.
2. Close other heavy applications. Whisper needs CPU/GPU headroom.
3. Free up RAM if usage is above 80%.
4. For very long recordings, split them into shorter sessions to keep memory pressure manageable.

The summarisation model has a context window. Very long transcripts get chunked, and occasionally a chunk fails silently. Try:

• Open the meeting and press Regenerate summary — most partial-summary issues clear on a single retry.
• Switch to a larger summarisation model (Settings → Summarisation) for denser material.
• If you're using Pro cloud AI with your own API key, check the provider dashboard for rate-limit errors.

Recordings live as WAV files in your OS-specific application support directory. Transcripts and summaries live in the local SQLite database. Nothing is uploaded by default.

• macOS: ~/Library/Application Support/NeuroBridgeEDU/
• Windows: %APPDATA%\NeuroBridgeEDU\
• Linux: ~/.local/share/NeuroBridgeEDU/

Yes. Recording, Whisper transcription, and Built-in AI summaries all work offline on every platform after the model download. Study Sage also works offline with Built-in AI. Ollama (local) and optional Pro cloud are additional provider choices.

Whisper supports many transcription languages on every platform. EuroLLM 9B is the Built-in AI option for European language mixing and code-switching; Mistral models are strongest on French and major European languages, and Built-in AI runs on-device on macOS, Windows, and Linux. The interface is available in 32 European languages, selectable in Settings.

Each platform handles loopback differently:

• macOS: Install BlackHole 2ch (free) and create a Multi-Output Device in Audio MIDI Setup. Pick BlackHole from the System Audio dropdown in the app.
• Windows: Pick the system audio source from the System Audio dropdown. Restart the app if it doesn't pick up.
• Linux: Verify your loopback source exists with pactl list sources (PulseAudio) or pw-cli ls Node (PipeWire), then select it from the dropdown.

Yes. The full source is on GitHub at EmminiX/NeuroBridge-EDU-OpenSource. The licence is non-commercial open source — usable freely by individuals and educational institutions.

Open a GitHub issue on the repo above (preferred — keeps the conversation public and discoverable for other users), or email contact@neurobridgeedu.eu if the issue includes anything sensitive.

Yes. The desktop app and this website are both built to WCAG 2.2 AA. Reduced motion is respected via prefers-reduced-motion. Screen reader semantics use native HTML landmarks (no DIY ARIA widgets). The app also ships a low-stim mode that desaturates colour and disables ambient motion entirely.