Talking to Omniscio and hearing it answer

What it is

This area is Omniscio's spoken conversation system — the parts that let you speak a request and hear a spoken reply. It is built in layers, each a little more capable than the last, and each with its own daily spending cap so a stray microphone or a long talk session cannot run up an unbounded bill.

• Report Back speaks a voice command's answer aloud: when you say something the command parser passes to the language model, the model's prose reply is read back through your chosen voice, not just shown as text.
• Layer 2, the general assistant holds a hands-free back-and-forth with Omniscio as a whole, not tied to any one session.
• Layer 3, session voice answers questions about a named session — its latest briefing, or a specific question — using only that thread's transcript, without opening it or interrupting it.
• Layer 4 carries a question to the coding agent running inside a thread and tells you when it has answered.

Voice history is recorded along the way, so you can go back and read what was said in any of these conversations.

Where to find it

• Settings → Voice Control — a top-level section with three accordion cards: Voice input, Text-to-speech output, and Voice Report Back. The Report Back card holds its on/off switch, and its Advanced settings reveal a maximum-length knob and a daily cost cap.
• A voice command — press Alt+V, use the wake word, or press the Test voice commands button. The spoken reply and a [Speak more] button appear on the floating voice-result card.
• Voice History in the sidebar — the list of your past voice conversations.
• Report Back reads aloud through whichever text-to-speech provider you have picked for any other speak action; there is no separate provider choice for it.

How it behaves

• Voice Report Back is the feature that reads a spoken command's answer aloud. It is off by default, speaks only the language model's prose replies, clips long answers behind a Speak more button, stops at a daily cost cap, and lists what deliberately stays silent.
• Voice, TTS & Wake Word is the overview of the whole stack — reading messages aloud, dictating into the composer, turning spoken sentences into actions, and the wake word. It also covers the Layer 2 general assistant.
• Voice, TTS & Wake Word (part 2) covers the engines behind the stack: the wake-word and speech-to-text choices with their keys and costs, the session briefing, and the per-layer daily spending caps.
• Voice L3 - Ask About a Thread covers asking hands-free about a named session, including the bounded follow-up loop and the offer to hand an unanswerable question to the live agent.
• Voice L4 - Ask the Live Agent covers asking the running agent a question and being told when it has answered.
• Voice History covers the persisted list of your voice conversations and how deleting one erases its transcript.
• Spoken Narration is the neighbouring read-aloud feature that plays a short spoken recap of each finished agent reply.

Related

• Voice Report Back (TTS read-back of voice-command answers) — speaking a voice command's reply aloud, off by default.
• Voice, TTS & Wake Word (hands-free Omniscio) — the full voice stack overview and the general assistant.
• Voice, TTS & Wake Word (hands-free Omniscio) (part 2) — the engines, providers and daily caps behind the stack.
• Voice L3 - Ask About a Thread (session:converse) — asking about a named session hands-free.
• Voice L4 - Ask the Live Agent (session:ask-agent) — asking the running agent and hearing the answer.
• Voice History — the recorded list of your voice conversations.
• Spoken Narration — a short spoken recap of each finished agent reply.