This page describes Dictum 0.5.

Command line

Dictum: voice to text. Results go to stdout; logs and progress go to stderr.

Every command takes --help (-h) to print its own help, and dictum --version prints the version.

Global options

These options work on every command, before or after the command's name. The commands below list only their own options.

Option	Description
--config	Config file to use instead of the default location
--verbose, -v	Log details to stderr, including dictated text

dictum init

Set Dictum up: choose and download a speech model, cleanup and your words

Usage: dictum init [options]

Option	Description
--yes, -y	Do not ask: use the options given and keep or recommend everything else
--language	Language you dictate in: one code such as uk, several such as "uk, en" to pick from, or auto
--stt	Speech model id (see 'dictum models list')
--llm	Text cleanup: none, or the base URL of an OpenAI-compatible server, e.g. http://192.168.1.10:8080/v1
--llm-model	Model name on that server
--llm-api-key	API key for that server: env:VAR, secret:name, or the key itself (then kept in the OS secret store)
--vocab-file	Words to add to the profile, one per line
--autostart	Start Dictum at login: the desktop app when dictum-gui sits next to dictum, otherwise the daemon
--gpu	auto (default) or cpu Values: auto, cpu
--json	Print a JSON summary

dictum transcribe

Transcribe a WAV file (any rate or channel count)

Usage: dictum transcribe <file> [options]

Argument	Description
<file>	WAV file to transcribe, or - for stdin

Option	Description
--profile, -p	Profile id from the config (default: the first profile)
--stt	STT provider id from the config
--model, -m	Whisper model file (name in the models directory, or a path); needs no config
--lang, -l	Language code such as uk, a list such as "uk, en" to pick from, or auto
--pipeline	Comma-separated processor ids, replacing the profile's chain
--raw	Skip all processing
--json	Print a JSON object with text, language, provider and timings

dictum record

Record from the microphone, then transcribe. Enter or Ctrl+C stops, Esc cancels.

Usage: dictum record [options]

Option	Description
--duration	Stop after this long, e.g. 30s
--until-silence	Stop after this much silence following speech, e.g. 2s
--device, -d	Microphone name or id from 'dictum devices'
--out	Where the text goes: stdout, clipboard, or paste into the focused app Default: stdout Values: clipboard, paste, stdout
--dump-frames	Write level and spectrum frames to a JSON file (for the overlay prototype)
--save-audio	Also save the recording as a 16 kHz WAV file, e.g. to check the microphone
--profile, -p	Profile id from the config (default: the first profile)
--stt	STT provider id from the config
--model, -m	Whisper model file (name in the models directory, or a path); needs no config
--lang, -l	Language code such as uk, a list such as "uk, en" to pick from, or auto
--pipeline	Comma-separated processor ids, replacing the profile's chain
--raw	Skip all processing
--json	Print a JSON object with text, language, provider and timings

dictum process

Run text through processors, e.g. dictum transcribe memo.wav --raw | dictum process --pipeline cleanup

Usage: dictum process [<text>] [options]

Argument	Description
<text>	Text to process, or - (or nothing) to read stdin

Option	Description
--pipeline	Comma-separated processor ids (default: the profile's chain)
--profile, -p	Profile whose chain and language to use
--lang, -l	Language of the text, passed to LLM prompts
--json	Print a JSON object with the text and per-processor timings

dictum stt

List and test speech-to-text providers

Usage: dictum stt <command>

Command	Description
list	List configured STT providers
test	Check a provider and transcribe a sample with it

dictum stt list

List configured STT providers

Usage: dictum stt list

dictum stt test

Check a provider and transcribe a sample with it

Usage: dictum stt test <id> [options]

Argument	Description
<id>	Provider id

Option	Description
--file, -f	WAV file to transcribe instead of recording 5 seconds
--lang, -l	Language code, a list such as "uk, en" to pick from, or auto

dictum profiles

List and inspect profiles

Usage: dictum profiles <command>

Command	Description
list	List profiles
set-hotkey	Change a profile's hotkey; a running daemon picks it up at once
show	Show a profile with its provider and processors

dictum profiles list

List profiles

Usage: dictum profiles list

dictum profiles set-hotkey

Change a profile's hotkey; a running daemon picks it up at once

Usage: dictum profiles set-hotkey <id> <hotkey>

Argument	Description
<id>	Profile id
<hotkey>	Combination, e.g. "Ctrl+Alt+Space" or "Ctrl+Alt+`"

dictum profiles show

Show a profile with its provider and processors

Usage: dictum profiles show <id>

Argument	Description
<id>	Profile id

dictum vocab

Words, names and terms the recognizer should spell exactly, per profile

Usage: dictum vocab <command> [options]

Option	Description
--profile, -p	Profile id (default: the first profile)

--profile also works on every subcommand of dictum vocab.

Command	Description
list	Print the profile's words, one per line
add	Add words to the end of the list
remove	Remove words from the list
clear	Remove every word from the list
import	Add words from a file

dictum vocab list

Print the profile's words, one per line

Usage: dictum vocab list

dictum vocab add

Add words to the end of the list

Usage: dictum vocab add <terms>...

Argument	Description
<terms>	Words or phrases; quote phrases with spaces

dictum vocab remove

Remove words from the list

Usage: dictum vocab remove <terms>...

Argument	Description
<terms>	Words or phrases to remove (any letter case)

dictum vocab clear

Remove every word from the list

Usage: dictum vocab clear

dictum vocab import

Add words from a file

Usage: dictum vocab import <file>

Argument	Description
<file>	Text file with one word or phrase per line ('#' starts a comment), or - for stdin

dictum models

List and download the speech models Dictum knows about

Usage: dictum models <command>

Command	Description
list	Show catalog models, whether they are installed, and other model files in the models directory
pull	Download models into the models directory (resumes after interruptions, checks SHA-256)

dictum models list

Show catalog models, whether they are installed, and other model files in the models directory

Usage: dictum models list [options]

Option	Description
--json	Print JSON
--all	Include models meant for tests

dictum models pull

Download models into the models directory (resumes after interruptions, checks SHA-256)

Usage: dictum models pull <ids>...

Argument	Description
<ids>	Catalog ids, e.g. whisper-small-q8

dictum devices

List microphones

Usage: dictum devices [options]

Option	Description
--json	Print JSON

dictum config

Create, locate, validate and edit the config file

Usage: dictum config <command>

Command	Description
path	Print the config file path
validate	Check the config for errors
edit	Open the config in $VISUAL / $EDITOR (or the system editor)

dictum config path

Print the config file path

Usage: dictum config path

dictum config validate

Check the config for errors

Usage: dictum config validate

dictum config edit

Open the config in $VISUAL / $EDITOR (or the system editor)

Usage: dictum config edit

dictum history

Show recent transcriptions

Usage: dictum history [options]

Option	Description
--last, -n	How many entries Default: 10
--json	Print JSON

dictum doctor

Check config, microphone and every provider, with hints for anything broken

Usage: dictum doctor [options]

Option	Description
--json	Print JSON

dictum daemon

Run in the background: hotkeys, dictation and the control socket (Ctrl+C stops)

Usage: dictum daemon [options]

Option	Description
--install-autostart	Start the daemon at login, then exit
--uninstall-autostart	Remove every Dictum login entry (the desktop app's or the daemon's), then exit

dictum toggle

Start dictation, or stop it and paste the text (bind this in your compositor)

Usage: dictum toggle [options]

Option	Description
--profile, -p	Profile id (default: the active or first profile)

dictum start

Start dictation

Usage: dictum start [options]

Option	Description
--profile, -p	Profile id (default: the active or first profile)

dictum stop

Stop dictation and paste the text

Usage: dictum stop [options]

Option	Description
--profile, -p	Profile id (default: the active or first profile)

dictum cancel

Stop dictation and discard it

Usage: dictum cancel [options]

Option	Description
--profile, -p	Profile id (default: the active or first profile)

dictum reload

Re-read the config file now

Usage: dictum reload [options]

Option	Description
--profile, -p	Profile id (default: the active or first profile)

dictum shutdown

Stop the running daemon

Usage: dictum shutdown [options]

Option	Description
--profile, -p	Profile id (default: the active or first profile)

dictum status

Show whether the daemon runs, its state, hotkeys and backends

Usage: dictum status [options]

Option	Description
--json	Print JSON
--watch	Keep printing state changes, results and status as JSON lines until Ctrl+C
--frames	With --watch, also print audio frames (50 per second)

dictum secret

Store API keys in the OS secret store, for "apiKey": "secret:<name>"

Usage: dictum secret <command>

Command	Description
set	Store a secret; the value is read from stdin or typed without echo
check	Tell whether a secret is stored, without printing it
delete	Remove a stored secret

dictum secret set

Store a secret; the value is read from stdin or typed without echo

Usage: dictum secret set <name>

Argument	Description
<name>	Secret name, e.g. groq

dictum secret check

Tell whether a secret is stored, without printing it

Usage: dictum secret check <name>

Argument	Description
<name>	Secret name

dictum secret delete

Remove a stored secret

Usage: dictum secret delete <name>

Argument	Description
<name>	Secret name