LIVECORE / LOCAL SELF-HOSTED STUDIO

Talk. Tune. Test.

A hands-on speech lab for personas, expressive stock voices, consented voice cloning, directedness, and speaker-state review.

Checking local stack…Gateway and data services
Development privacy boundary Microphone turns are processed ephemerally. Raw conversation audio is not retained. Transcripts default to seven days. A custom voice sample stays only in gateway memory and is deleted on revoke or restart.

01 / SESSION

Shape the conversation

DISCONNECTED
Locked

Loading persona catalog…

Loading environment modes…

Connect to grant microphone permission and start a local WebRTC session.

02 / LIVE TURN

Conversation

—round trip
No turns yetConnect, then hold the talk button while speaking.
This checks controller behavior without using your microphone or the live speech models.

03 / VOICE STUDIO

Enroll a voice clone

Explicit consent required

Create a temporary local voice profile from an audio sample you own. This is separate from identifying who is speaking.

No custom voices enrolled in this gateway process.

04 / SPEAKER LAB

Speaker labels & identity

Checking capability…
1
Speaker 1ANONYMOUS · session label only
Live biometric identification is not enabled. The bundled speaker service is a deterministic contract fixture and must not be used to identify people.
Why voice cloning and identification are separate

A voice clone generates speech from an authorized reference. Speaker identification compares an embedding against consented profiles. LiveCore never treats either one as authentication.

AMBIENT SHADOW

Low-trust room context

Ambient text cannot authorize tools, memory, identity, policy changes, or confirmations.

Empty · low trust · expires automatically