Creating custom voices

A custom voice gives your AI SDR agent a voice trained from real recordings, so visitors hear a voice that fits your brand in voice and video conversations. After processing, the custom voice appears in your agent profile alongside the standard voices.

What you need

You need two recordings from the person whose voice you're using:

• Consent recording: the speaker reads the consent statement shown on screen, confirming that they own the voice and agree to its use for your agent.
• Voice sample: about 30 seconds of the speaker reading the passage shown on screen. This is the recording your custom voice is trained on.

Both recordings can be in any of these formats: MP3, WAV, OGG, AAC, FLAC, WebM or MP4. Most phone and laptop voice memo apps save to one of them.

Note: The consent recording must come from the person whose voice is in the voice sample. Someone else, such as a colleague or an executive, can't give consent on the speaker's behalf.

Recording guidelines

Your custom voice reproduces the tone, pace, pauses and accent of the voice sample exactly as recorded, and it can't be edited after it's created. Before you record:

• Record in a quiet space with minimal echo. Soft furnishings help, while bare or tiled rooms add reverb to the voice.
• Keep the microphone 7 to 8 inches from the speaker. An external microphone works well at that distance.
• Keep energy, style and accent consistent through the whole sample.
• Read naturally, the way you want your agent to sound in a live conversation.
• Record several takes, listen back and upload the best one.

How to create a custom voice

1. Navigate to Agent Studio → Agent Profile.
2. Click + Create a custom voice under the Voice dropdown.
3. Name the voice. The name is what you'll see in the voice dropdown, so choose one that identifies it, for example "Sarah, US English".
4. Upload the consent recording.
5. Upload the voice sample.
6. Click Create voice.

The consent recording and the voice sample use different scripts, so check that each file is in the right upload before you click Create voice.

Processing takes some time, so your custom voice may not appear in the voice dropdown right away. Once it's ready, select it in the voice dropdown to assign it to your agent, then click Preview to hear it. For more on choosing and previewing voices, see Setting up your agent profile in Agent Studio.

Important considerations

• A custom voice can't be edited after it's created. To change how it sounds, record new takes and create a new voice.
• Your agent's voice and persona are set separately. To give your agent a custom face as well, see Creating custom personas.
• A custom voice pairs well with the Mode-specific guidance and Commonly misheard terms fields in your Conversations guide. See Configuring Guides in Agent Studio.
• You are responsible for securing the consent of the person whose voice you use.

Custom voice limits