Package Exports

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (dikt) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

dikt

Voice dictation for the terminal. Record, transcribe, copy — zero npm dependencies.

Uses Mistral's Voxtral for speech-to-text.

Install

npm install -g dikt

Requires sox for audio recording (not needed for --file):

# macOS
brew install sox

# Ubuntu/Debian
sudo apt install sox

# Arch
sudo pacman -S sox

Optional dependencies for --file mode:

ffmpeg — enables compression, chunked transcription of long files, and broader format support
yt-dlp — enables transcribing audio from URLs (YouTube, podcasts, etc.)

Setup

On first run, dikt will prompt you for your Mistral API key and model preferences:

dikt setup

Config is stored in ~/.config/dikt/config.json.

Usage

dikt

This opens an interactive TUI where you can record, transcribe, and copy text.

Keys

Key	Action
`Space`	Start / stop recording
`c` / `Enter`	Copy transcript to clipboard
`a`	Toggle auto-copy
`h`	Cycle through history
`r`	Re-transcribe last recording
`Esc`	Cancel recording
`s`	Re-run setup
`?`	Show keybindings
`q`	Quit

Single-shot mode

# Print transcript to stdout
dikt -q

# Output JSON
dikt --json

# Pipe to another tool
dikt -q | claude

# Wait longer before auto-stopping
dikt -q --silence 5

Stream mode

Continuously transcribe, emitting chunks on pauses:

dikt --stream

# Stream as JSON Lines
dikt --stream --json

# Stream as continuous flowing text
dikt --stream -n

# Stream continuously until Ctrl+C
dikt --stream --silence 0

File mode

Transcribe an existing audio file (wav, mp3, m4a, flac, ogg, webm, aac, wma, and more):

dikt --file meeting.wav

# Save to a file (.json auto-enables JSON output)
dikt --file meeting.wav -o transcript.json
dikt --file meeting.wav -o transcript.txt

# With JSON output
dikt --file recording.mp3 --json

# Transcribe from a URL (requires yt-dlp)
dikt --file https://youtube.com/watch?v=VIDEO_ID
dikt --file https://youtube.com/watch?v=VIDEO_ID -o transcript.txt

Speaker identification & timestamps

# Speaker labels
dikt -q --diarize

# Timestamps
dikt -q --timestamps segment
dikt -q --timestamps word
dikt --file lecture.mp3 --timestamps segment

# Combined with JSON
dikt -q --json --diarize

Options

Flag	Description
`--file <path\|url>`	Transcribe audio file or URL (via yt-dlp)
`-o`, `--output <path>`	Write output to file (`.json` auto-enables JSON)
`--stream`	Stream transcription chunks on pauses
`--json`	Output JSON (single-shot or stream)
`-q`, `--quiet`	Record once, print transcript to stdout
`--silence <seconds>`	Silence duration before auto-stop (default: 2.0)
`--pause <seconds>`	Pause duration to split stream chunks (default: 1.0)
`--language <code>`	Language code, e.g. en, de, fr (default: auto)
`--timestamps <granularity>`	Add timestamps: segment or word
`--diarize`	Enable speaker identification
`-n`, `--no-newline`	Join stream chunks without newlines
`--no-color`	Disable colored output
`--no-input`	Fail if config is missing (no wizard)
`--setup`	Run setup wizard
`--update`	Update to latest version
`--version`	Show version
`-h`, `--help`	Show help

Update

dikt update

Environment variables

Variable	Description
`DIKT_API_KEY`	Override API key
`DIKT_MODEL`	Override model (default: `voxtral-mini-latest`)
`DIKT_LANGUAGE`	Override language (default: auto)
`DIKT_TEMPERATURE`	Override temperature
`DIKT_CONTEXT_BIAS`	Override context bias

License

MIT