Voice Mode

Voice Mode lets end users hold a real-time, hands-free conversation with your chatbot Flow or Agent. They talk; it transcribes, thinks, and speaks back - with no buttons to press while talking.

It's a great fit for:

• Phone-like concierges - bookings, walk-throughs, support.
• Accessibility - users who would rather listen than read.
• Hands-busy use cases - training in the field, kitchen instructions, in-car assistants.

---

Where Voice Mode works

Voice Mode is built for chatbot Flows and Agents. Form Flows (which take structured input fields) don't use voice.

When voice is enabled and the model produces an artifact - like a generated document or chart - end users still see it on screen. Voice is for the conversational layer, not for replacing visual output.

---

Enabling Voice Mode

Voice Mode is configured per Agent - either the Agent product or the Agent node in a chatbot Flow.

For an Agent

1. Open your Agent in the Agents section.
2. Go to the Personality step, then the Voice sub-tab.
3. Toggle Conversational voice on.
4. Pick a voice from the voice picker.
5. Publish your Agent to make voice available to end users.

For a chatbot Flow

1. Open your Flow in the workflow builder.
2. Click the Agent node where you want voice enabled.
3. In the agent settings panel, find the Voice section and toggle it on.
4. Pick a voice.

[Screenshot: Voice toggle and voice picker inside the Agent builder]

---

Choosing a voice

The voice picker is split into two groups:

• Your voices - Voices you've cloned or generated yourself. These let you build a branded sound.
• Library - A curated set of pre-made voices covering a range of styles and accents.

In both groups you can:

• Search by name.
• Filter by gender, accent, or category.
• Preview a voice with the play button before picking it.

Heads up - The voice picker depends on an ElevenLabs connection. If you see an error or no voices, head to Settings > Integrations and connect ElevenLabs first.

How Vibe shapes the voice

If you've set a Vibe on your Agent (in Personality > Vibe), FormWise nudges the selected voice's delivery to match. Roughly:

• Playful or casual - more expressive style.
• Serious or formal - steadier, more consistent delivery.
• Creative - a little more variation between turns.

You don't have to tune anything yourself - pick a vibe and a voice, and FormWise handles the blending.

---

What end users experience

When an end user opens a voice-enabled chat:

• They tap the voice button to start a call. There's no push-to-talk - it listens continuously.
• They see their words appear as a live transcript while they talk.
• The Agent speaks the answer back. As it speaks, a teleprompter lights up the words in time with the audio (karaoke-style). They can hide the teleprompter with an eye toggle if they want a pure-audio experience.
• Tap the orb to interrupt. End users don't have to wait for the Agent to finish - a tap stops playback and lets them speak again. This barge-in feels natural in real conversations.
• A mute button silences their mic; a red end-call button closes the voice session.
• The mic-volume "halo" around the end button gives a visual cue that it can hear them.

[Screenshot: Voice mode active with live transcript and teleprompter]

---

Tips

• Match voice to vibe. A "warm" vibe with a clipped, businesslike voice feels off. Preview before publishing.
• Keep answers short. Voice answers should sound like answers, not essays. Add a guardrail like "Keep voice responses under three sentences unless the user asks for detail."
• Test on mobile. Most end users will use voice on a phone. Try it on your own device before sharing.
• Plan for noisy rooms. Set expectations in your persona ("If I mishear you, I'll ask again.") rather than fighting it.

Limitations

• Voice needs a browser microphone permission. End users will see a permission prompt the first time they start a call.
• Voice features depend on a connected ElevenLabs account.
• Voice does not run inside Form Flows - only chatbot Flows and Agents.
• Cloned voices follow ElevenLabs' usage rules. Make sure you have rights to any voice you clone.

Next steps