Commands — LLMVoice Docs

Diagnostics

`llmvoice doctor`

Inspect Python, FFmpeg, PyTorch, CUDA, model access, disk space, and writable paths.

llmvoice doctor
llmvoice doctor --debug

Voice commands

`llmvoice voice add`

Create or update a named local voice from reference audio.

llmvoice voice add friday reference.wav

`llmvoice voice list`

List every registered voice.

llmvoice voice list

`llmvoice voice info`

Show the stored path and audio metadata for a voice.

llmvoice voice info friday

`llmvoice voice remove`

Remove a voice from the local registry. This does not touch unrelated source recordings.

llmvoice voice remove friday

Configuration commands

Use show for the effective configuration, get for one value, and set to persist a default.

llmvoice config show
llmvoice config get default_voice
llmvoice config set default_voice friday

Generate speech

`llmvoice start`

Generate a single MP3 from a text transcript.

llmvoice start transcript.txt
llmvoice start transcript.txt --voice friday

Long input is split into chunks, synthesized sequentially, merged, and encoded. The terminal reports the chosen engine, device, estimated duration, progress, and output summary.

Global flags

--dry-run validates inputs and prints the planned run without loading XTTS-v2.
--force allows an existing output file to be replaced.
--debug prints additional diagnostic context and stack information.

llmvoice start transcript.txt --dry-run
llmvoice start transcript.txt --force
llmvoice start transcript.txt --debug