SimplyDictate

Voice-to-clipboard dictation via OpenAI Whisper (or any compatible transcription API), built for Niri (or similar WMs) and Wayland.

Bind a single key to toggle recording: press once to start, press again to stop. The audio is transcribed and the text lands on your clipboard, ready to paste. A second binding cancels an in-progress recording. If music is playing, it is paused while you record and resumed afterwards.

How it works

Recording toggles on each invocation, tracked via a PID file in $XDG_RUNTIME_DIR/dictate/.
Audio is captured with parecord (mono, 16 kHz WAV).
On stop, the WAV is POSTed to the transcription API and the result is piped into wl-copy.
Desktop notifications report state (recording, transcribing, copied, errors).

Requirements

System dependencies:

Tool	Package	Purpose
`parecord`	`pulseaudio-utils`	Audio capture (works via pipewire-pulse)
`wl-copy`	`wl-clipboard`	Copy to the Wayland clipboard
`notify-send`	`libnotify`	Desktop notifications (optional)
`playerctl`	`playerctl`	Auto-pause/resume media (optional)

Python dependency: requests.

Installation

git clone https://github.com/Coffei/simply_dictate.git ~/.local/share/dictate
ln -s ~/.local/share/dictate/run.sh ~/.local/bin/dictate

Create a secrets file (mode 600, kept outside the repo) with your API key:

mkdir -p ~/.config/dictate
install -m 600 /dev/null ~/.config/dictate/env
$EDITOR ~/.config/dictate/env

# ~/.config/dictate/env
OPENAI_API_KEY=sk-...

run.sh refuses to load this file unless it is mode 600. It also prefers a project virtualenv at .venv/ if present, otherwise falls back to system python3.

Niri binding

In ~/.config/niri/config.kdl:

binds {
    Mod+V       { spawn "dictate"; }
    Mod+Shift+V { spawn "dictate" "cancel"; }
}

Configuration

All configuration is via environment variables (set them in ~/.config/dictate/env):

Variable	Default	Description
`OPENAI_API_KEY`	(required)	API key for the transcription service.
`DICTATE_MODEL`	`gpt-4o-transcribe`	Transcription model.
`DICTATE_LANGUAGE`	`""`	ISO language code; empty enables auto-detect.
`DICTATE_PROMPT`	(tech vocabulary)	Vocabulary bias (≤ ~224 tokens / ~900 chars).
`DICTATE_API_URL`	OpenAI transcriptions endpoint	Override to use a compatible API (see below).
`DICTATE_KEEP_AUDIO`	(unset)	Set to `1`/`true`/`yes` to save timestamped recordings for debug.
`DICTATE_ENV_FILE`	`~/.config/dictate/env`	Override the secrets file location.

Using Groq instead of OpenAI

DICTATE_API_URL=https://api.groq.com/openai/v1/audio/transcriptions
DICTATE_MODEL=whisper-large-v3-turbo

License

MIT

Coffei/simply_dictate

README