VELZYX LAB talk loop
LAB only, not production speech-to-speech. The brain is a scripted LAB STUB: not Sovereign, not an LLM, and it hears no words. It picks the next receptionist line from turn count and turn length only. Talk once replies with the Speak placeholder's non-speech hum: no words, no voice, and nothing from your audio goes into it (no cloning). Live converse speaks the LAB STUB line as English words from the VelzyxFormantV0 rule synthesizer: a robotic synthetic voice, not modelled on anyone, not a Speak model, and nothing from your audio goes into it. Turn is the energy baseline (-40 dBFS). VELZYX Listen, Turn, and Speak models: NOT STARTED. Latency: not measured.