No SSML audio tag is rendered in Google Assistant main invocation

Question

I have the following invocation:

candidates:
  - first_simple:
      variants:
        - speech: >-
            <speak><audio
            url="https://example.com/welcome.mp3"></audio></speak>
          text: Welcome to my App!
    suggestions:
      - title: Explore
      - title: Guess

When I test it I do not hear mp3 being played, only see welcome text.

What I do incorrectly here?

Answer 1

It turns out that using SSML is only possible in a webhook. Returning speech from a webhook resolved the issue.