With Dialogflow CX, you can let virtual agents respond with pre-recorded audio. This lets you use high-quality audio files instead of standard text-to-speech. This capability is available for all voice channels, including inbound and outbound calls. It's available for support virtual agents, task virtual agents, and post-session virtual agents. Barge-in is not supported during pre-recorded audio playback.
For more information, see Play prerecorded audio.
Before you begin
Host your audio files in a publicly-accessible service such as Cloud Storage. The audio file must be in WAV format.
Configure pre-recorded audio for virtual agent responses
To configure pre-recorded audio for virtual agent responses, follow these steps:
Go to the Conversational Agents console.
Select the project that contains your agent.
Select your agent.
In the navigation menu, click Flows.
Click the flow that you want.
In the flow pane, click the page that you want.
For Entry fulfillment, click Edit Fulfillment. The Fulfillment dialog appears.
Click Add dialogue response > Play pre-recorded audio.
In the Play pre-recorded audio field, enter the full, publicly-accessible URI for your WAV file—for example,
gs://YOUR_BUCKET/YOUR_AUDIO_FILE.wav.Optional: Add additional dialog response types. You can use the Move up and Move down arrows to reorder the dialog responses.
Click Save.
Failure handling
If playback of pre-recorded audio fails (due to an inaccessible URI or a malformed audio file, for example), you can expect the following behavior:
The system silently skips that fulfillment segment so that the end-user isn't aware of the failure.
Other fulfillments in the same response are processed as expected.
The system logs an error that includes the following:
Timestamp
Session ID
Virtual agent ID
Error code or error message