Skip to main content

Voice & Transcription

Capture ideas by speaking. Telemore's voice features give you a system-wide speech-to-text shortcut and meeting transcription.

Global Speech-to-Text

Press Alt+Z (or Option+Z on Mac) anywhere to start recording. Press again to stop. Your speech is transcribed via Whisper and the result is pasted at your cursor.

How It Works

  1. Recording — the WhisperRecorderManager captures audio from your microphone with real-time audio-level visualization
  2. Transcription — audio is sent to the Whisper service for server-side transcription
  3. Output — the transcribed text is inserted at your cursor position via the Tauri clipboard bridge

Status Indicators

StateVisual
Readymic icon (idle)
Initializingspinner
Recordingpulsing mic icon with audio level bars
Transcribingloading indicator with "Transcribing..." text

Settings

Configure in Settings → Account → Transcription:

SettingDefaultDescription
Enabledoffmaster toggle for the transcription feature
Auto-pasteoffautomatically paste transcription into the active app
ShortcutAlt+Zthe global hotkey (not currently user-configurable)
Promptemptyoptional context hint to improve transcription accuracy

Settings are persisted to the server via the employee update endpoint and cached in localStorage.

Meeting Transcription

During a meeting, enable Record Meeting to capture system audio. Audio is buffered, segmented, and transcribed in real time using the Whisper service. The transcript appears in the live transcript panel with:

  • Timestamp per chunk
  • Confidence score
  • Cumulative cost estimate
  • Speaker text

After the meeting ends, the full transcript is sent to the AI for summary generation. See Meetings for details.

Chat Input Voice

The ChatInput component includes a mic button for voice input. Click the mic icon to record, then the transcribed text is appended to your message. This works in:

  • AI chat sidebar
  • Team chat message input
  • Any ChatInput instance

Device Capabilities

The useDeviceCapabilities hook detects whether the current device has a microphone. If no mic is detected, voice features are hidden or disabled with an appropriate message.