Pickaxe Learn

Changelog
New featureBuild ·

Pick the voice your agent speaks in

Try in the builder

Create an agent or open an existing agent in the builder, then follow the steps below. Model and feature availability may depend on your plan.

Pick the voice your agent speaks in

What it does

Read-out was rebuilt around a voice picker: search a library of natural-sounding voices in your workspace's language, preview any of them, and — if you would rather use your own — connect an ElevenLabs account instead.

Before this

An expansion, and a large one. Read-out already existed as a switch, but there was no voice to choose — every agent that spoke used the same one.

Why it matters

The voice is most of the impression an audio answer makes, and until now every Pickaxe agent that spoke made the same one.

How it works

Allow users to hear responses used to be a switch with nothing behind it — every agent that could speak used the same voice. It now has a Choose voice button next to it in the builder's Configure tab, opening a picker with two tabs.

Pickaxe voices
The built-in library. Step through the voices or search by name, accent, or gender, and play a sample of any of them before you commit. Each card shows the voice's gender and locale.
Language
The list follows your workspace's language rather than defaulting to English — 27 are mapped, from Arabic to Zulu, each with that language's own voices.
Custom voice
Added on 21 August. Connect an ElevenLabs account with an API key and pick from the voices on it, including ones you have cloned.
Voice ID
Optional, on the Custom voice tab. Leave it blank to browse your ElevenLabs voice list, or paste an exact Voice ID to go straight to one voice.

The choice is stored on the agent, so it travels with the agent to every deployment rather than being set per place it is embedded.

Built-in voices cost nothing extra and need no account. A connected ElevenLabs key needs Text to Speech, Voices, and Voice Generation permissions — a key missing one of the three fails when it tries to list voices — and the audio it generates is billed by ElevenLabs rather than drawn from your Pickaxe credits.