hyprvox

Press a key. Speak. Paste.
Voice-to-text for Linux, built for AI workflows.

Why

You're deep in a session with a coding agent. You know exactly what to say — a complex refactor, a debugging question, a long feature spec. But now you have to type it all out.

By the time you finish, the thought is gone.

Typing at 40 WPM when you speak at 150 WPM is a bottleneck you don't need. hyprvox closes that gap.

How It Works

hyprvox is a background daemon. Hold a key, speak, release — your words land on the clipboard, ready to paste into Claude, Copilot, or any agent.

Under the hood:

| Engine | Model | Role | |--------|-------|------| | Groq | Whisper V3 | Speed | | Deepgram | Nova-3 | Accuracy | | LLM merge | Llama 3.3 70B | Best-of-both result |

Both engines run in parallel. Results merge with an LLM. If one fails, the other carries on. The merged text is validated before it ever reaches your clipboard: hallucinated outros are trimmed, prompt artifacts and reasoning leakage are blocked, injected filename/command bursts are rejected, and technical terms are preserved.

Median latency: 882ms. 39× faster than real-time.

Quick Start

1 — Install system dependencies

# Fedora / RHEL
sudo dnf install -y unzip ffmpeg alsa-utils wl-clipboard libnotify

# Arch
sudo pacman -S ffmpeg alsa-utils wl-clipboard libnotify

# Ubuntu / Debian
sudo apt install ffmpeg alsa-utils unzip wl-clipboard libnotify-bin

On X11, install xclip instead of wl-clipboard.

2 — Install Bun

curl -fsSL https://bun.sh/install | bash
export BUN_INSTALL="$HOME/.bun"
export PATH="$BUN_INSTALL/bin:$PATH"

Confirm Bun is available:

bun --version

3 — Install hyprvox

Use the published CLI for normal installs:

bun add -g hyprvox
hyprvox --help

Or run from a source checkout:

git clone https://github.com/Snehit70/hyprvox.git
cd hyprvox
bun install
bun run index.ts --help

4 — Configure

Run the setup wizard:

hyprvox setup

The wizard detects your distro, desktop session, missing commands, config state, microphone, Wayland hotkey strategy, and systemd service state. It keeps reporting the remaining next step until setup is complete.

For source checkouts, use:

bun run index.ts setup

For Docker, CI, or non-interactive diagnosis:

hyprvox setup --check --json

Manual config-only setup is still available:

hyprvox config init   # Enter your Groq + Deepgram API keys

Source checkout equivalent: bun run index.ts config init.

| Key | Get one at | |-----|-----------| | Groq | console.groq.com | | Deepgram | console.deepgram.com |

5 — Verify the setup before installing the service

hyprvox setup --check
hyprvox list-mics
hyprvox health

Source checkout equivalents:

bun run index.ts setup --check
bun run index.ts list-mics
bun run index.ts health

health should confirm the config, API keys, microphone, clipboard command, and notification support. On headless servers, clipboard, notification, and microphone checks can fail by design.

To change config later, run the interactive editor:

hyprvox config

For direct commands and scripting, run hyprvox config --help.

6 — Install and start the user service

Check whether the daemon is already running before installing the service:

hyprvox status

If it is already running manually, stop it first:

hyprvox stop

Then install the systemd user service:

hyprvox install   # Installs as a systemd user service
hyprvox status

Source checkout equivalents: bun run index.ts install and bun run index.ts status.

Keybinds

Hyprland / Wayland

Add to ~/.config/hypr/hyprland.conf:

bind = , code:105, exec, hyprvox toggle
# code:105 = Right Control — use `wev | grep -A5 "key event"` to find others

X11 / GNOME / KDE

The built-in Right Ctrl hotkey works out of the box.

The Overlay

A waveform appears at the bottom of your screen while recording — visible confirmation it's listening.

For Hyprland, add to ~/.config/hypr/UserConfigs/WindowRules.conf:

windowrule = match:class hyprvox-overlay, float on
windowrule = match:class hyprvox-overlay, pin on
windowrule = match:class hyprvox-overlay, no_focus on
windowrule = match:class hyprvox-overlay, no_shadow on
windowrule = match:class hyprvox-overlay, no_anim on
windowrule = match:class hyprvox-overlay, move ((monitor_w-window_w)*0.5) (monitor_h-window_h-50)

Commands

hyprvox toggle       # Start / stop recording
hyprvox status       # Daemon status
hyprvox health       # Full system check
hyprvox stats        # Interactive stats dashboard
hyprvox stats --summary  # Plain stats summary
hyprvox history      # Past transcriptions
hyprvox logs         # Tail daemon logs
hyprvox errors       # Last error
hyprvox boost add    # Add custom vocabulary
hyprvox config init  # Reconfigure API keys

Configuration

Config lives at ~/.config/hypr/vox/config.json:

{
  "apiKeys": { "groq": "...", "groqFallback": "...", "deepgram": "..." },
  "transcription": {
    "streaming": true,
    "formattingMode": "clean",
    "boostWords": ["Hyprland", "WebSocket", "refactor"]
  }
}

| Option | Values | Effect | |--------|--------|--------| | streaming | true / false | Stream Deepgram during recording for lower latency | | mergeModel | Groq model ID | LLM used for transcript merge; default is llama-3.3-70b-versatile | | formattingMode | verbatim clean structured | How the merged output is shaped | | boostWords | string list | Improves recognition of technical terms |

Full reference → docs/CONFIGURATION.md

Troubleshooting

| Symptom | Fix | |---------|-----| | Hotkey not firing (Wayland) | Use compositor bind — see Hyprland Setup | | Hotkey not firing (X11) | Add user to input group, re-login | | No audio captured | sudo usermod -aG audio $USER then re-login | | Empty recording after restart | Run one more recording; if it repeats, test with arecord -D default -f S16_LE -r 16000 -c 1 -d 1 /tmp/mic.wav | | Clipboard not updating | Install wl-clipboard (Wayland) or xclip (X11) | | Service won't start | journalctl --user -u hyprvox -f |

Headless servers (EC2, etc.): health checks for clipboard, notifications, and mic will fail by design. The install is valid if hyprvox --help works — full transcription needs a machine with audio input.

Full guide → docs/TROUBLESHOOTING.md

For AI Agents

Install and configure hyprvox on this Linux system:

1. Install deps: sudo dnf install -y unzip ffmpeg alsa-utils wl-clipboard libnotify
2. Install Bun: curl -fsSL https://bun.sh/install | bash && export BUN_INSTALL="$HOME/.bun" && export PATH="$BUN_INSTALL/bin:$PATH"
3. Clone: git clone https://github.com/Snehit70/hyprvox.git && cd hyprvox && bun install
4. Configure: bun run index.ts config init  (I'll provide Groq + Deepgram keys when prompted)
5. Run the setup wizard: bun run index.ts setup
6. Verify non-interactively: bun run index.ts setup --check
7. If needed, install service manually: bun run index.ts install
8. For Hyprland, add to ~/.config/hypr/hyprland.conf:
     bind = , code:105, exec, bun run /path/to/hyprvox/index.ts toggle
9. Add overlay window rules to ~/.config/hypr/UserConfigs/WindowRules.conf (see README)
10. Reload: hyprctl reload
11. Verify: bun run index.ts status && bun run index.ts health

Docs

• Architecture — internals and data flow
• STT Flow — transcription and quality pipeline end-to-end
• Configuration — every option
• CLI Commands — every command and flag
• Wayland Support — platform-specific setup

License

MIT