CLI

The wh command controls the Local Whisper service and exposes local utilities for recording, transcription, text-to-speech, history, diagnostics, updates, and configuration.

Service lifecycle

wh                  # Status and help
wh status           # Service status, PID, grammar backend, uptime, and recovery state
wh start            # Launch the service
wh stop             # Stop the service
wh restart          # Restart and rebuild Swift UI if sources changed
wh build            # Rebuild Swift UI app
wh log              # Tail service log
wh version          # Show version
wh uninstall        # Remove Local Whisper

Engines and grammar

wh engine                 # Show current engine and list available engines
wh engine parakeet_v3     # Switch to Parakeet-TDT v3
wh engine qwen3_asr       # Switch to Qwen3-ASR
wh engine qwen3_asr 1.7b  # Higher-quality default
wh engine qwen3_asr 0.6b  # Lower memory and latency
wh engine apple_speech    # Switch to Apple SpeechTranscriber
wh engine whisperkit      # Switch to WhisperKit

wh backend                # Show current grammar backend and list available backends
wh backend apple_intelligence
wh backend ollama
wh backend lm_studio

The CLI switch restarts the service. The macOS menu bar can switch engines and grammar backends in place. Choose Qwen3-ASR 1.7B (higher-quality default) or 0.6B (lower memory and latency) directly with wh engine qwen3_asr <size>, or use wh config. The selected MLX snapshot is downloaded before activation. wh config also controls Qwen vocabulary. When enabled together with Vocabulary replacements, the validated Qwen models receive those rules as local context/hotwords (up to 4,096 characters). Add and manage the underlying word list with wh replace; unsupported engines receive none of this data.

Recording and transcription

wh listen                 # Record until silence, then print transcription
wh listen 30              # Record up to 30 seconds
wh listen --raw           # Raw transcription, no grammar

wh transcribe recording.wav
wh transcribe --raw audio.wav

Text-to-speech

wh whisper "hello"
wh whisper --voice af_bella "hello"
echo "hello" | wh whisper

Text-to-speech uses Kokoro MLX in process after the model is installed. No local TTS server is required.

Replacements

wh replace
wh replace add "gonna" "going to"
wh replace remove "gonna"
wh replace on
wh replace off
wh replace import rules.csv
wh replace export rules.csv
wh replace test "gonna say this"

Imports accept CSV, TSV, TOML-style "spoken" = "replacement", or arrow-style spoken -> replacement files, and land in one atomic write. Replacement edits hot-reload a running service — no restart needed. wh replace test shows what the current rules would produce for a sample sentence.

History and diagnostics

wh stats
wh export
wh export --format json --out ~/Downloads/history.json
wh export --format txt --limit 50

wh config
wh config edit
wh config path
wh doctor
wh doctor --fix
wh doctor --report
wh update

Diagnostic reports redact local transcript and config contents. They are designed to be safe to include in private support or issue reports.

Start

Product

Reference

Service lifecycle

Engines and grammar

Recording and transcription

Text-to-speech

Replacements

History and diagnostics

​Service lifecycle

​Engines and grammar

​Recording and transcription

​Text-to-speech

​Replacements

​History and diagnostics

Service lifecycle

Engines and grammar

Recording and transcription

Text-to-speech

Replacements

History and diagnostics