Generate speech

Before you start

Complete the quickstart. Use a voice available to your account and check model-specific controls before making a paid request.

kitta audio models list --json
kitta audio voices list --all --language en --json
kitta audio tts --schema

Text or file

kitta audio tts --voice VOICE_ID --text "Hello from Kitta" --output hello.mp3
kitta audio tts --voice VOICE_ID --text-file script.txt --format wav --output narration.wav

Save script.txt as UTF-8. Replace VOICE_ID with a discovered voice ID. Use a different output path for each completed file; the CLI never overwrites an existing file. --output waits for completion. Without it, use --wait if you want to wait without saving.

Standard input

Bash:

printf '%s' 'Hello from Kitta' | kitta audio tts --voice VOICE_ID --text - --output stdin.mp3

PowerShell:

$OutputEncoding = [System.Text.UTF8Encoding]::new()
"Hello from Kitta" | kitta audio tts --voice VOICE_ID --text - --output stdin.mp3

PowerShell may append a newline. Keep the exact input unchanged when recovering a submission.

Controls and result

Use --model with an ID returned by models list; model and voice compatibility is enforced by the server. Consult --schema for format and control limits. --no-text-normalization explicitly disables normalization. A successful output is a fully downloaded audio file. Diagnostics and task IDs go to stderr.

If waiting times out, do not start a new generation just to obtain the file. Continue with recovery and billing.