I've developed a quite simple skills that returns response with the audio tag in ssml. Since yesterday, the alexa rendering of the responses (which are text + audio + text) breaks in the middle of the audio rendering.
Strange facts are:
1. The audio files are correctly converted according to the ask documentation, the are long about 3 min, so they do not exceed the limit of 240 secs.
2. The response in rendered correctly both on the alexa simulator and on mobile devices, the problemas seems to be only on echo devices (I've tested with 2 echo dots, not on other devices unfortunatly)
Here an example of the response that does not work.
<p>Oggi vi parlo di Maurice Wilkins, il terzo uomo del DNA</p> <audio src="https://firebasestorage.googleapis.com/v0/b/fs-admin-395d4.appspot.com/o/9uTQUgzvG4YWJlhyblMV%2Faudio.mp3?alt=media&token=669eae84-d2ed-4917-ba54-66a95d028311" /> <p>Grazie per aver ascoltato questa skill. Aprimi domani per ascoltare un nuovo contenuto!</p>
The audio is hosted on google firebase storage, not on S3, but it's should not be an issue since other audio files works and anyway the audio file is correclty downloaded by the devices.