Speak for Me

Speak for Me lets you type what you want to say and have your agent speak it aloud in the meeting, word for word, using its configured ElevenLabs voice. It's available on all plans and designed for anyone who can't or prefers not to participate verbally — including people with speech difficulties, voice fatigue, or anyone in a noisy environment.

Open your live meeting →

How it works#

When a meeting is live, a Speak for Me panel appears on the meeting page below the agent instructions. You type your message, choose whether to use the Hand Raise Protocol, and send. Your agent speaks with its own voice and persona, exactly as if it had triggered naturally.

The spoken text also appears in the meeting chat alongside the audio, so participants can follow along in text as well.

Two modes#

Direct#

Type exactly what you want your agent to say. The text is passed through speech normalisation (the same processing applied to all agent responses) and injected as audio immediately.

Use Direct when you know precisely what you want to say and don't need any help composing it.

AI-Assisted#

Type rough notes describing what you want to communicate — a few words or a sentence is enough. Your agent uses Claude to expand those notes into a clear, natural spoken response that fits the agent's persona and the meeting context.

Before anything is spoken, you see the generated response in an editable text area. You can revise it as much as you like, then confirm to speak it.

Use AI-Assisted when you have an idea but want help shaping it into something well-phrased and natural.

Hand Raise Protocol#

Both modes include a Hand Raise toggle, which defaults to on.

Hand Raise on (default): Your agent raises its hand in the meeting — its camera tile shows a "Waiting to speak" indicator — and waits to be called on by name before speaking. This is the least disruptive option and gives you full control over when the agent's voice enters the conversation.

Hand Raise off: Your agent speaks immediately when you send the message, without waiting. Use this when the timing is already right and you don't want the delay of being called on.

The toggle is shared between Direct and AI-Assisted tabs — switching tabs preserves your choice.

Character limit#

Each message can be up to 2,000 characters. A counter below the text area shows how many characters you have remaining. The AI-Assisted generated response is also capped at 2,000 characters before it is sent to TTS.

AI-Assisted call limits#

The AI-Assisted mode uses Claude to generate responses, so it has a per-meeting cap based on your plan:

Direct mode has no limit — you can type and speak as many times as you like in a meeting regardless of plan.

View plans and pricing →

Availability#

Speak for Me is available on all plans, including Free. It appears on the meeting page only when the meeting status is Live — it is not available during the joining or waiting room phase because the bot is not yet in the call.

Tips#

• Keep it conversational. Your agent's voice and speech normalisation are tuned for natural spoken language. Bullet points, symbols, and shorthand don't always translate well to audio — write as you would say it.
• Use AI-Assisted for longer contributions. If you want to make a substantive point, give the AI-Assisted mode a sentence of context and let it compose the full response. You can trim or rewrite the result before speaking it.
• Hand Raise is better in formal settings. In client calls, presentations, or any structured meeting, using Hand Raise means the agent never interrupts — it signals it has something to say and waits for the right moment.
• Your agent's persona applies. In AI-Assisted mode, the generated response is shaped by the agent's configured persona prompt. If your agent is configured as a sales coach, the response will sound like one.