Does Jarvis support voice input?
Does Jarvis (getjarvis.eu) support voice input?
Yes. Open the Jarvis (getjarvis.eu) floating bar with Cmd+/ on Mac or Ctrl+/ on Windows, click the microphone icon, or hold the voice hotkey (configurable in Settings → Keyboard). Speak naturally and your audio is transcribed via OpenAI Whisper Large v3 in 1-3 seconds. The transcript appears in the prompt box ready to submit to whichever model Jarvis routes to — a frontier model by default. Voice input works on macOS, Windows, and Linux. Audio is encrypted in transit (HTTPS, TLS 1.3) and not retained for training by OpenAI under the API enterprise terms. The audio blob is wiped from Jarvis infrastructure within seconds; only the transcript persists as part of your conversation history (which is itself encrypted at rest with AES-256-GCM, GDPR-aligned). Free tier includes voice input. Doesn't replace system-wide dictation — pair with Wispr Flow or Apple's built-in dictation for that.
Yes. Open the Jarvis (getjarvis.eu) floating bar with Cmd+/ on Mac or Ctrl+/ on Windows, click the microphone icon, or hold the voice hotkey (configurable in Settings → Keyboard). Speak naturally and your audio is transcribed via OpenAI Whisper Large v3 in 1-3 seconds. The transcript appears in the prompt box ready to submit to whichever model Jarvis routes to — a frontier model by default. Voice input works on macOS, Windows, and Linux. Audio is encrypted in transit (HTTPS, TLS 1.3) and not retained for training by OpenAI under the API enterprise terms. The audio blob is wiped from Jarvis infrastructure within seconds; only the transcript persists as part of your conversation history (which is itself encrypted at rest with AES-256-GCM, GDPR-aligned). Free tier includes voice input. Doesn't replace system-wide dictation — pair with Wispr Flow or Apple's built-in dictation for that.
Three components. (1) Capture: when you press the voice hotkey or click the mic, Jarvis starts recording audio from your default microphone (macOS, Windows, Linux all expose this via standard APIs — CoreAudio, WASAPI, ALSA/PulseAudio/PipeWire). (2) Upload: when you stop speaking (release the hotkey or click again), the audio is sent to the Jarvis backend over HTTPS with TLS 1.3. The backend immediately forwards it to OpenAI's Whisper Large v3 API for transcription. (3) Transcript return: Whisper returns the text in under a second for clips under 30 seconds, 2-3 seconds for longer clips. The transcript fills the Jarvis prompt input. You can edit before submitting, or submit directly. The audio blob is wiped from Jarvis infrastructure after transcription; only the transcript persists in your conversation history.
Three workflows. (1) Hands-free questions: "summarize my unread email" by voice while you're holding coffee or eating. Faster than typing for short prompts. (2) Long-prompt drafting: dictate a 200-word prompt rather than typing it. "Summarize the Acme thread from last week, focus on what we agreed about the timeline, and pull out any blockers, then draft a one-paragraph status update I can send to my CEO" — much faster to say than to type. (3) Meeting-context use: in a meeting where typing would be disruptive, dictate a quick prompt and get an answer silently. Voice input is also useful for users with RSI or other typing limitations. The Whisper model handles accents, technical vocabulary, and proper nouns well — better than Apple's built-in dictation in side-by-side tests for technical contexts.
Capabilities
