Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.hyprcore.ai/llms.txt

Use this file to discover all available pages before exploring further.

This is the five-minute path from finishing the installer to your first transcribed sentence.
1

Open any text field

Slack, Notes, your terminal, an email draft—anywhere your cursor blinks.
2

Press your dictation shortcut

The default depends on your platform; you set it during onboarding. Open Settings → Dictation to see or change it.The recording overlay appears at the top or bottom of your screen with a live waveform.
3

Speak naturally

Hyprcore uses voice activity detection, so you don’t need to clip the start or end. Pause as long as you want; the model only transcribes speech.
4

Press the shortcut again

Hyprcore stops recording, runs the transcription locally, and pastes the result wherever your cursor was.
Prefer push-to-talk? Turn on Push to talk in Settings → Dictation. Hold the shortcut to record, release to transcribe.

​
What just happened

  1. Hyprcore captured audio from your microphone.
  2. Silero VAD filtered silence so the model only sees real speech.
  3. Your selected speech model—Parakeet V3 by default—transcribed the audio on-device.
  4. If you have post-processing turned on, an LLM cleaned up filler words, fixed casing, and tightened the wording.
  5. Hyprcore pasted the final text into the app that had focus when you started.
No audio left your Mac. No transcript left your Mac. Hyprcore stored a copy in your local dictation history so you can reuse it later.

​
Try a meeting

Once dictation works, record a meeting. Same idea—shortcut, speak, get text—except you also get screen video, speaker labels, an editable summary, and a permanent wiki page.

​
If something didn’t work

  • No text appeared. Check Accessibility permission in System Settings, then quit and relaunch Hyprcore.
  • Audio was empty. Check Microphone permission, and confirm the right input device in Settings → Audio.
  • Shortcut didn’t fire. Check Input Monitoring permission. On Wayland Linux, see Troubleshooting.
  • Transcript was wrong. Try a larger model (Whisper Medium → Turbo) or a different one—see Choosing a speech model.