​
Installation

The CLI is installed automatically with Izwi. Verify installation: 
izwi --version

​
Global Options

These options work with all commands:
OptionDescription
--server <URL>Server URL (default: http://localhost:8080)
--config <PATH>Configuration file path
--output-format <FORMAT>Output format: table, json, plain, yaml
--quietSuppress all output except results
--verboseEnable verbose output
--no-colorDisable colored output
--helpShow help information
--versionShow version

​
Commands

​
Server

CommandDescription
serveStart the inference server
statusShow server health and status

​
Models

CommandDescription
listList available models
pullDownload a model
rmRemove a downloaded model
modelsModel management subcommands

​
Inference

CommandDescription
ttsText-to-speech generation
transcribeSpeech-to-text transcription
diarizeSpeaker diarization
alignForced alignment
chatInteractive chat

​
Utilities

CommandDescription
benchRun benchmarks
configManage configuration
completionsGenerate shell completions
versionShow version information

​
Quick Examples

​
Start the server

izwi serve
izwi serve --mode desktop
izwi serve --port 9000

​
Download and use models

izwi list
izwi pull Qwen3-TTS-12Hz-0.6B-Base
izwi list --local

​
Generate speech

izwi tts "Hello world" --output hello.wav
izwi tts "Hello world" --play

​
Transcribe audio

izwi transcribe audio.wav
izwi transcribe audio.wav --format json

​
Speaker diarization

izwi diarize meeting.wav
izwi diarize meeting.wav --num-speakers 3

​
Forced alignment

izwi align audio.wav "Hello world, this is a test."

​
Interactive chat

izwi chat
izwi chat --model Gemma-3-1b-it --system "You are a helpful assistant"

​
Getting Help

Get help for any command: 
izwi --help
izwi serve --help
izwi tts --help

​
Environment Variables

VariableDescription
IZWI_HOSTServer host
IZWI_PORTServer port
IZWI_MODELS_DIRModels directory
IZWI_BACKENDBackend preference (auto, cpu, metal, cuda)
IZWI_MAX_BATCH_SIZEMaximum batch size
IZWI_MAX_CONCURRENTMaximum concurrent requests
IZWI_TIMEOUTRequest timeout (seconds)
RUST_LOGLog level
IZWI_LOG_FORMATLog output format (text, json)
NO_COLORDisable colored output

​
See Also

Server

izwi serve

Start the Izwi inference server and configure host, port, backends, logging, and runtime behavior.

izwi status

Check Izwi server health, model readiness, runtime status, and detailed diagnostics.

izwi config

Manage Izwi configuration values, files, and precedence from the command line.

izwi version

Show Izwi version, build information, and compiled backend support.

izwi completions

Generate shell completions for Bash, Zsh, Fish, PowerShell, and Elvish.

Models

izwi list

List available and downloaded Izwi models with status and storage details.

izwi pull

Download an Izwi model to the local model cache.

izwi rm

Remove a downloaded Izwi model from local storage.

izwi models

Use model management subcommands to list, inspect, load, unload, and track model progress.

Workflows

izwi tts

Generate text-to-speech audio from the command line with model, voice, and output options.

izwi transcribe

Transcribe audio files from the command line with output formats and ASR model options.

izwi diarize

Run speaker diarization from the command line and export text or JSON results.

izwi align

Align reference text to audio and export timestamped word or segment results.

izwi chat

Start an interactive local chat session from the command line.

Benchmarks

izwi bench

Run chat, TTS, ASR, throughput, manifest, and regression comparison benchmarks.