Skip to main content
Vocalis uses the OpenAI Realtime API to power the Actor’s voice. This is what makes the AI sound natural and conversational — speech goes directly in and out without a text intermediary.

Prerequisites

  • An OpenAI account with API access (platform.openai.com)
  • Access to the Realtime API (may require a usage tier upgrade)

Step 1: Create an API key

1

Go to API keys

Log in to platform.openai.com and navigate to API keys.
2

Create a new key

Click Create new secret key.
  • Name: vocalis-production (or similar)
  • Permissions: Ensure it has access to the Realtime API models
3

Copy the key

Copy the key immediately — you won’t be able to see it again.
Store your API key securely. Do not share it or commit it to version control.

Step 2: Verify Realtime API access

The Realtime API is required for Vocalis to function. To verify you have access:
  1. Check your OpenAI usage tier. The Realtime API may require Tier 1 or above.
  2. Ensure your account has billing set up with sufficient credits.

Step 3: Connect to Vocalis

1

Open Vocalis settings

Log in to vocalis.witting.ai and go to SettingsIntegrations.
2

Enter your API key

In the OpenAI section, paste your API key.
3

Select a default model

Choose the default model for your Actors. Vocalis supports OpenAI Realtime models — select the latest available option.
4

Save and test

Click Save. Use the Test Connection button to verify the key works.

Voice options

When configuring an Actor, you can select from the available OpenAI voices. Each voice has a different tone and character:
VoiceDescription
alloyNeutral, balanced
echoWarm, conversational
fableExpressive, storytelling
onyxDeep, authoritative
novaFriendly, upbeat
shimmerClear, professional
Test different voices with your system prompt to find the best match for your Actor’s persona. A technical recruiter might suit nova or alloy, while an executive reference check might work better with onyx.

Usage and costs

Realtime API usage is billed by OpenAI based on audio duration. Monitor your usage at platform.openai.com/usage. Typical interview costs depend on call duration:
  • A 15-minute phone screen uses approximately 15 minutes of audio input + output
  • Set OpenAI usage limits to avoid unexpected charges

Troubleshooting

IssueSolution
”Invalid API key”Regenerate the key and re-enter it. Ensure no whitespace was copied.
”Model not available”Check your OpenAI usage tier. The Realtime API may require a tier upgrade.
Voice sounds roboticEnsure you’re using a Realtime API model, not a standard completion model.

Next steps