NeuralScio includes voice support for spoken queries, dictated notes, and spoken playback inside the research workflow.
Available Voice Backends
| Backend | Status | Notes | |----------|--------|-------| | OpenAI Realtime | Implemented | Cloud voice path | | Gemini Live | Implemented | Cloud voice path | | Grok Voice | Implemented | Cloud voice path | | Local voice path | Implemented | Speech-to-text plus text-to-speech components on the local machine |
What Voice Is Good For
- dictating a quick research question
- capturing notes without leaving the current reading context
- listening to summaries or passages
- accessibility and hands-occupied workflows
Local Voice Path
The local voice path is the strongest privacy-oriented setup. Desktop code shows:
- Whisper-based speech-to-text support
- local model integration
- system text-to-speech with optional Piper support
Voice works best as a supporting layer for dictation and playback, not as a replacement for the main desktop UI.
Cloud Voice Path
The cloud backends are useful when you want:
- lower-latency conversational behavior
- managed provider-quality speech handling
- voice sessions tied to the same cloud provider stack as the rest of your AI setup
Setup Guidance
- Open the voice settings in the desktop app.
- Choose a cloud or local backend.
- Add the required provider credentials for cloud backends, or configure local components for the local path.
- Test microphone and playback before relying on voice in a longer session.
Privacy Boundary
- Cloud voice sends audio to the selected provider.
- Local voice keeps processing on your machine.
- The best setup depends on whether you prioritize convenience or local control.