CLI Reference

🌐 Language: English | Français

CLI Reference

Complete reference for every command shipped by dictee. All commands are installed under /usr/bin/ (system) or ~/.local/bin/ (user install).

Three categories: shell entrypoints (dictee, dictee-setup, dictee-switch-backend), Python helpers (dictee-tray, dictee-postprocess, dictee-test-rules), and Rust binaries (transcribe, transcribe-daemon, transcribe-client, transcribe-diarize, transcribe-stream-diarize, transcribe-diarize-batch).

Shell entrypoints

`dictee`

Main entrypoint. Toggles dictation on/off, or runs one-shot modes via flags.

Synopsis:

dictee [--translate] [--meeting] [--cancel] [--setup] [--reset]

Flags:

Flag	Description
(none)	Toggle dictation — press once to start, press again to stop
`--translate`	Transcribe + translate to `$DICTEE_LANG_TARGET`
`--meeting`	Transcribe with diarization (speaker labels)
`--cancel`	Cancel the current recording
`--setup`	Launch `dictee-setup` (equivalent to running setup wizard)
`--reset`	Clear state files in `/dev/shm` — use if dictee is stuck
`--help`	Print help

Examples:

# Plain dictation
dictee

# Translate FR → EN
DICTEE_LANG_SOURCE=fr DICTEE_LANG_TARGET=en dictee --translate

# Meeting mode (diarization)
dictee --meeting

# Cancel ongoing recording
dictee --cancel

`dictee-setup`

PyQt6 GUI for full configuration. Launches the setup wizard on first run, or opens the main configuration panel on subsequent runs.

Synopsis:

dictee-setup [--wizard] [--tab=<name>] [--lang=<xx>]

Flags:

Flag	Description
(none)	Open main config panel
`--wizard`	Force wizard mode (even if already configured)
`--tab=<name>`	Jump to tab: `asr`, `translation`, `postprocess`, `rules`, `shortcuts`, `plasmoid`, `advanced`
`--lang=<xx>`	Start with UI locale `xx` (overrides `$LANG`)

`dictee-switch-backend`

Switch ASR or translation backend without restart.

Synopsis:

dictee-switch-backend <subcommand> [args]

Subcommands:

Command	Description
`status`	Show current ASR + translation backends
`list`	List all available backends
`asr <name>`	Switch ASR to `parakeet` / `canary` / `whisper` / `vosk`
`translate <name>`	Switch translation to `canary` / `libretranslate` / `ollama` / `google` / `bing`

Examples:

dictee-switch-backend status
# → ASR: parakeet (dictee.service, active)
# → Translate: google (trans)

dictee-switch-backend asr canary
dictee-switch-backend translate ollama

`dictee-ptt`

Push-to-talk daemon that listens to /dev/input/event* for keycodes and triggers dictation. Alternative to native KDE/GNOME shortcuts.

Synopsis:

dictee-ptt [--mode=toggle|hold|double-tap] [--key=<code>] [--key-translate=<code>] [--mod-translate=alt|ctrl|shift|super] [--double-tap-ms=<ms>]

See Keyboard-Shortcuts#dictee-ptt-daemon for detailed usage. Normally managed by the systemd user service dictee-ptt.service.

Python helpers

`dictee-tray`

System tray icon. See Tray-Icon.

Synopsis:

dictee-tray [--debug] [--no-autostart]

Normally started via the dictee-tray.service systemd user unit.

`dictee-postprocess`

Runs the 12-step post-processing pipeline on stdin or a given text, outputs the result on stdout.

Synopsis:

dictee-postprocess --lang <xx> [--llm-position=first|last|hybrid|off]
                   [--no-translate] [--rules=<path>] [--dict=<path>]

Examples:

echo "hello comma how are you" | dictee-postprocess --lang en
# → "hello, how are you"

# Disable LLM for a test
echo "rough asr output" | dictee-postprocess --lang en --llm-position=off

# Custom rules file
echo "text" | dictee-postprocess --lang en --rules ~/my-rules.conf

`dictee-test-rules`

Interactive test harness for regex rules. See Rules-and-Dictionary#testing-rules-live.

Synopsis:

dictee-test-rules [--loop] [--wav <file>] [--rules <path>] [--lang <xx>]

Examples:

# Interactive — type text, see output
dictee-test-rules --lang fr

# Continuous loop for rapid iteration
dictee-test-rules --loop

# Test a real audio file
dictee-test-rules --wav recording.wav

Rust binaries

`transcribe`

One-shot CLI transcription of a WAV file. Not used by the dictation flow (which uses the daemon) but handy for batch transcription.

Synopsis:

transcribe <audio.wav> [--model <path>] [--lang <xx>]

Daemon / client

transcribe-daemon holds the model in memory and listens on a Unix socket. transcribe-client connects to the daemon and sends audio/requests.

Daemon synopsis:

transcribe-daemon [--parakeet|--canary|--nemotron] [--model <path>] [--lang <xx>]

Client synopsis:

transcribe-client <audio.wav>            # transcribe a file
transcribe-client --record               # record from mic and transcribe

Socket path: $XDG_RUNTIME_DIR/transcribe.sock (defaults to /run/user/$UID/transcribe.sock).

`transcribe-diarize`

Sortformer + Parakeet-TDT in one call. Transcribes + assigns speaker labels.

Synopsis:

transcribe-diarize <audio.wav> \
  --sortformer-model <path> \
  --parakeet-model <path> \
  [--lang <xx>] [--format=plain|rttm|json|srt]

Caveats: the single-shot binary hits a ~5:20 min hard cap on every GPU due to a known bug in the Parakeet-TDT v3 model. For long files, use transcribe-diarize-batch (see below) or dictee-transcribe (the GUI), which auto-routes through the chunked binary since v1.3.4. More context: Diarization#combined-transcription-mode.

`transcribe-stream-diarize`

Sortformer + Nemotron (EN only) in streaming mode. No duration cap.

Synopsis:

transcribe-stream-diarize <audio.wav> \
  --sortformer-model <path> \
  --nemotron-model <path> \
  [--format=plain|rttm|json|srt]

`transcribe-diarize-batch`

Chunked pipeline for long audio with diarization. Runs Sortformer once globally, splits the audio, transcribes each chunk with Parakeet, then merges speaker labels via timestamps. Each chunk stays under the ~5:20 min Parakeet bug, so audio of any length works (54-min keynote diarized in 122 s on 8 GB).

Synopsis:

transcribe-diarize-batch <audio.wav> \
  --sortformer-model <path> \
  --parakeet-model <path> \
  --chunk-sec=600 --overlap-sec=10 \
  [--lang <xx>] [--format=plain|rttm|json|srt]

Since v1.3.4, dictee-transcribe (the GUI) wires this binary automatically for any file long enough to benefit from chunking — no manual invocation needed.

Environment variables

Runtime configuration. Set via dictee.conf or exported in your shell:

Dictation mode

Var	Values	Default	Description
`DICTEE_ASR_BACKEND`	`parakeet`, `canary`, `whisper`, `vosk`	`parakeet`	Active ASR backend
`DICTEE_LANG_SOURCE`	ISO 639-1	`$LANG`	Source language
`DICTEE_LANG_TARGET`	ISO 639-1	`en`	Translation target
`DICTEE_TRANSLATE`	`0` / `1`	`0`	Never persisted — ephemeral only
`DICTEE_MEETING`	`0` / `1`	`0`	Meeting mode (diarization)

Post-processing

Var	Values	Description
`DICTEE_PP_LLM`	`0` / `1`	Disable LLM correction for this dictation
`DICTEE_PP_NUMBERS`	`0` / `1`	Toggle number conversion
`DICTEE_LLM_MIN_WORDS`	integer	Skip LLM for dictations shorter than N words

Backends

Var	Default	Description
`DICTEE_OLLAMA_HOST`	`http://localhost:11434`	Ollama endpoint
`DICTEE_OLLAMA_MODEL`	`gemma3:4b`	Ollama model tag
`DICTEE_LIBRETRANSLATE_HOST`	`http://localhost:5000`	LibreTranslate endpoint
`DICTEE_WHISPER_MODEL`	`large-v3-turbo`	faster-whisper model
`DICTEE_VOSK_MODEL`	(auto)	Vosk model path

Runtime

Var	Default	Description
`DICTEE_FORCE_CPU`	`0`	Force CPU even if GPU is available
`DICTEE_INTRA_THREADS`	(auto)	ONNX Runtime intra-op thread count (default: `min(8, nproc)`). Lower (2-4) to save battery on laptops or share CPU with other workloads. Do not exceed your CPU thread count.
`DICTEE_PARAKEET_QUANT`	(auto)	Parakeet model variant: `int8` (~670 MB, ~34 % faster on CPU with AVX-VNNI) or `fp32` (~2.4 GB, fastest on GPU). Both variants can coexist on disk; this var selects which one is loaded. Default: auto-detected from VRAM (int8 if < 4 GB or CPU-only, else fp32).
`DICTEE_DEBUG`	`0`	Verbose logging
`DICTEE_STATE_DIR`	`/dev/shm`	Override state file location

Exit codes

Code	Meaning
`0`	Success
`1`	Generic error
`2`	Invalid arguments
`3`	Daemon not running / connection refused
`4`	Model file missing
`5`	Audio capture failed
`6`	Transcription failed
`7`	Translation failed
`8`	Post-processing error
`126`	Permission denied (e.g., `/dev/input` without `input` group)

State files

Runtime state lives in /dev/shm (tmpfs, cleared on reboot). Multi-user safe via UID suffix.

File	Purpose
`/dev/shm/.dictee_state_<UID>`	Current status: `idle` / `recording` / `transcribing` / `offline`
`/dev/shm/.dictee_state_<UID>.lock`	flock coordination for concurrent writes
`/dev/shm/.dictee_toggles_<UID>`	LLM / Short / Meeting toggle state
`/dev/shm/.dictee_continuation_<UID>`	Last-word buffer for continuation feature
`$XDG_RUNTIME_DIR/transcribe.sock`	Daemon Unix socket
`~/.config/dictee.conf`	Persistent user configuration
`~/.config/dictee/rules.conf`	User regex rules (merged on top of system defaults)
`~/.config/dictee/dictionary.conf`	User dictionary
`~/.config/dictee/short_text_keepcaps.conf`	User keepcaps exceptions

Next steps

Troubleshooting — diagnose CLI errors
Developer-Guide — binary source architecture
Keyboard-Shortcuts — dictee-ptt in depth
Post-Processing-Overview — what dictee-postprocess does

📖 dictee Wiki

🇬🇧 Home · 🇫🇷 Accueil

Getting started / Premiers pas

Installation · 🇬🇧 · 🇫🇷
Setup-Wizard · 🇬🇧 · 🇫🇷
Configuration · 🇬🇧 · 🇫🇷
Plasmoid-Widget · 🇬🇧 · 🇫🇷
Tray-Icon · 🇬🇧 · 🇫🇷
Keyboard-Shortcuts · 🇬🇧 · 🇫🇷
Voice-Commands · 🇬🇧 · 🇫🇷
GPU-Setup · 🇬🇧 · 🇫🇷
Diarization · 🇬🇧 · 🇫🇷
LLM-Diarization · 🇬🇧 · 🇫🇷

Speech recognition / ASR

ASR-Backends · 🇬🇧 · 🇫🇷
Parakeet-TDT-Deep-Dive · 🇬🇧 · 🇫🇷
Canary-1B-Deep-Dive · 🇬🇧 · 🇫🇷

Translation / Traduction

Translation · 🇬🇧 · 🇫🇷
Ollama-Setup · 🇬🇧 · 🇫🇷

Post-processing / Post-traitement

Overview · 🇬🇧 · 🇫🇷
Rules-and-Dictionary · 🇬🇧 · 🇫🇷
LLM-Correction · 🇬🇧 · 🇫🇷
Numbers-Dates-Continuation · 🇬🇧 · 🇫🇷

CLI

CLI-Reference · 🇬🇧 · 🇫🇷

Reference / Référence

Troubleshooting · 🇬🇧 · 🇫🇷
FAQ · 🇬🇧 · 🇫🇷
Developer-Guide · 🇬🇧 · 🇫🇷
Changelog · 🇬🇧 · 🇫🇷

🏠 Repo · 📦 Releases · 🐛 Issues

CLI Reference

CLI Reference

Table of Contents

Shell entrypoints

dictee

dictee-setup

dictee-switch-backend

dictee-ptt

Python helpers

dictee-tray

dictee-postprocess

dictee-test-rules

Rust binaries

transcribe

Daemon / client

transcribe-diarize

transcribe-stream-diarize

transcribe-diarize-batch

Environment variables

Dictation mode

Post-processing

Backends

Runtime

Exit codes

State files

Next steps

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

`dictee`

`dictee-setup`

`dictee-switch-backend`

`dictee-ptt`

`dictee-tray`

`dictee-postprocess`

`dictee-test-rules`

`transcribe`

`transcribe-diarize`

`transcribe-stream-diarize`

`transcribe-diarize-batch`