Docs / Configuration

Configuration

Set reliable defaults for each generation run.

View configuration

Inspect every effective value or retrieve a single key:

llmvoice config show
llmvoice config get default_voice

Core settings

  • default_voice — registered voice used when --voice is omitted
  • default_language — XTTS language code, such as en
  • default_speed — speech speed multiplier
  • device — automatic selection, cuda, or cpu
llmvoice config set default_voice friday
llmvoice config set default_language en
llmvoice config set default_speed 1.0
llmvoice config set device cuda

Generation settings

  • chunk_size controls the target amount of text sent through each synthesis pass.
  • chunk_pause_ms controls the silence inserted between merged chunks.

Smaller chunks can reduce recovery cost when a run fails, but may create more boundaries. Pause length should suit the cadence of the source transcript.

llmvoice config set chunk_size 350
llmvoice config set chunk_pause_ms 180

Example setup

A practical local CUDA configuration:

llmvoice config set default_voice friday
llmvoice config set default_language en
llmvoice config set default_speed 1.0
llmvoice config set device cuda
llmvoice config set chunk_size 350
llmvoice config set chunk_pause_ms 180
llmvoice config show