Voice Systems · Provider Control

Pocket / Piper

System-wide speech routing for VPM, voice loop, and normal speech.

Remote control token

System-wide mode

Uses Pocket's bounded generation temperature; Piper fallback ignores this setting.
Saved mode
Active default
Preferred
Readiness
Actual
Fallback
Last requester

Provider health

Piper
Pocket
Pocket queue
Last error

Generation-only preview

Generation details (collapsed)
Loading provider status…

Generate preview creates a local WAV. It does not send audio to a physical Lantern.

The system-wide default is used only when the selected Lantern has no per-device voice assignment. A per-Lantern assignment intentionally overrides this setting.

Kyutai online Pocket voice library

Browse the metadata-only catalog from kyutai/tts-voices, select one voice, preview its reference recording, and download only that voice. Refreshing the catalog never downloads the library.

Catalog not loaded.

Select a voice to inspect its local state.

Voice libraries

Load the Piper catalog or Pocket’s built-in voices, then choose a voice for a local generation-only sample. Piper installs require the control token and download model files to Core.

Piper voice loader

Browse the official Piper catalog, filter by language or quality, and install one model at a time.

Pocket voice library

Pocket voices are bundled with the running service. Loading this list refreshes the available built-in voice inventory.

Uses the generation-only sampler below; it never sends audio to Lantern.

Provider voice sampler

This sampler always uses /api/audio/tts/test and the local audio player. It never calls the Lantern playback path.

Choose a provider and voice.