Skip to main content
Google Cloud provides the voice and transcription, using one service-account JSON for both.

​
Getting a service-account JSON

  1. In the Google Cloud Console, select or create a project.
  2. Enable the APIs you need — Cloud Text-to-Speech for voice, Cloud Speech-to-Text for transcription.
  3. Go to IAM & Admin → Service Accounts and create one.
  4. Open its Keys tab, choose Add Key → Create new key, and select JSON.
  5. The file downloads once. Store it securely and treat it as a secret.
Choose JSON, not P12. P12 is a legacy format kept for older Google Workspace integrations and will not work here.
Upload the file as-is. It must be the whole JSON, including private_key and client_email — do not extract individual values from it.

​
Adding the connection

  1. Go to Connections and choose Google Cloud.
  2. Use Upload JSON to attach the service-account file.
  3. Choose which services this connection drives — TTS, STT, or both.
  4. Save.

​
Using separate service accounts for voice and transcription

One connection can drive both services, but they do not have to share a credential. To keep them apart — separate projects, quotas, or regions — add a second Google Cloud connection with its own label and service account, then pick which connection each service uses on Model Configurations.

​
What you can select afterwards

Choose your models on Model Configurations. Voice — pick from the dropdown, or enter a fully-qualified Google Cloud voice ID rather than a short name. Google’s voice list has samples. Language — Google Cloud uses BCP-47 codes, such as ar-XA and cmn-CN. Set it to match your callers. See Google’s supported languages.

​
Troubleshooting

Confirm the matching API is enabled on the project — Cloud Text-to-Speech for voice, Cloud Speech-to-Text for transcription. A valid service account for a project without the API enabled fails at call time.
Google Cloud expects its own fully-qualified voice identifiers and BCP-47 language codes. Copy the exact ID from Google’s voice list rather than typing a short name.
Make sure it is the JSON file, not P12, and that you pasted the whole file contents rather than a single field from it.
An AI Studio API key belongs on the Gemini connection and drives the language model. Google Cloud speech does not accept one.