Agent voices

Your leads can speak, and you can speak to them

Typing is always there, and voice is an extra. It shows up in four places:

Voice needs your workspace's voice service to be set up. Where it is not, the controls say so, for example "Voice is not set up on this workspace yet.", and typing carries on as usual.

Each lead has a voice of its own

Every team lead speaks in its own voice, picked to suit its role, and no two leads share one. When several leads answer in the same room, you can tell who is talking without looking. A lead keeps the same voice from one sentence to the next.

The leads with a voice are MIRA, NIA, MOXA, HOOK, Echo, SERA, NORA, DEXA, RIVA, CREA, RUPA, ALPHA, DRIS, CORA, LIPI and VANI.

Read aloud in a team room (Beta)

Every team room has Read aloud at the top right, in the row with Huddle. Press it, and from then on each new message from an agent is spoken in that agent's voice. Its note says: "New messages are read aloud, each in its lead's voice. The room does not listen."

• Only what arrives next. The history already on screen is not read back.
• Only agents. Your own words, your colleagues' words and system notes are never spoken.
• Speaking only. The room never uses your microphone. Several leads may answer in a room, so you reply by typing, or by dictating (below).
• Off until you press it. Sound never starts on its own, and Read aloud is off again when you leave the room.

Press Mute voice to stop, and Unmute voice to carry on. While muted, new messages stay on screen. Unmuting does not read back what came in meanwhile.

If your browser holds the sound back, the note says "The browser held the sound back. Press Unmute voice to hear it."

For Prompt World's Marketing room: press Read aloud, type "@mira what did we learn from last week's workshop sign-ups?", and MIRA's answer is read to you as it arrives. See The team room.

Talk to your lead in the set-up chat (Beta)

The set-up chat can run hands-free. Press Turn voice on. Your lead then speaks its replies as they are written and listens to you. The note names your lead, for example "SERA speaks its replies and listens. Talk, or type as usual." What you say is sent as if you had typed it. While you talk, the note reads Hearing you.

• Mute and Unmute stop and restart it. Typing works either way.
• Spoken language appears once voice is on. It is English unless you pick Detect automatically, and your browser remembers the choice.
• Voice here needs the microphone. If your browser does not allow it, the chat says so and stays typed.

When a workspace admin chooses Set up your workspace on the welcome, the chat it opens starts with voice on, so the browser asks for sound and the microphone once, on that same press. See Setting up with your team lead.

Dictate into any chat (Beta)

Many chat boxes in the admin carry a small microphone button, Dictate: the team room composer, direct messages, the set-up dock and other chat panels.

1. Press the microphone and allow your browser to use it.
2. Speak. A timer runs while you record, for up to two minutes, and recording stops by itself at two minutes.
3. Press again to stop. Your words appear in the box.

Nothing is sent until you press send, so you can correct the text first. If voice input is not set up on your workspace, the button says "Voice input is not set up on this workspace." and then hides.

What voice costs

Voice is paid work, so it runs inside your budgets like any other. Each sentence a lead speaks is charged to that lead, and each dictated clip to the agent the chat talks to, against its budget, before the work is done. When the budget is spent, voice stops and says so: "Read aloud is off: its budget has run out." in a room, and "Voice is off: its budget has run out. Typing still works." in the set-up chat. See Credits and budgets.

Voices for video presenters

In Video Studio, every persona that can appear on screen has its own voice. A workspace admin sets it on a draft persona in Personas: click Complete, then pick the Voice engine and enter the Voice id at the engine. The engines are:

• Fish Audio
• Kinetiq voice
• ElevenLabs (Beta)

You can also make a voice under A voice made here (Beta):

• Design a voice. Type Describe a voice to design, for example "Warm, unhurried, a light Mumbai accent, mid thirties", optionally set Voice gender and Accent, as a locale, and click Design voices. Listen to each offer and click Use this voice.
• A digital twin's own voice. Clone their voice from a recording of them speaking, an MP3 or WAV in your workspace's library, or Use the voice from their footage. Both need a written release on file that covers their voice.

When a video is planned, the studio's Settings offer the voices the preset allows. See ALPHA, the video studio.

Not the same as a writing voice

An agent also has a writing voice: how it phrases what it writes. That is a prompt, not a sound. When an agent writes in the voice Nagent gives it, the agent workbench's Profile shows that prompt under Platform default voice. See The agent workbench. Your brand's written voice is set in Brand Lock.

Where to go next

• The team room: where Read aloud lives.
• Setting up with your team lead: the set-up chat.
• Credits and budgets: how voice is paid for.
• ALPHA, the video studio: presenters and their voices.