CODEWITHSUNDEEP
amazon-web-servicesamazon-s3aws-lambdaaws-api-gatewayalexa-skills-kit
Bailey
Bailey

Reputation: 167

Playing an mp3 created by AWS api gateway in an Alexa skill

I have an AWS lambda function that is called by AWS api gateway. The function takes URLs from multiple mp3 files hosted on AWS S3 and concatenates them into a single mp3 file. When I call the api from a browser all is good (the browser opens a media player and the combined audio mp3 is played).

The URL request looks like this:

https://0xxxxxxxx.execute-api.eu-west-1.amazonaws.com/alpha/files?file=https://xxx.s3-eu-west-1.amazonaws.com/file1.mp3&file=https://xxx.s3-eu-west-1.amazonaws.com/file2.mp3&file=https://xxx.s3-eu-west-1.amazonaws.com/file3.mp3

the HTTP response is of type audio/mpeg, about 10 seconds long, and is base64 encoded.

I've tried to wrap this into SSML in my skill and it fails. From the Alexa skills kit voice and tone simulator, I get the error message "error retrieving text to speech; the input was incompatible"

In the simulator, this is what I wrote:

<speak> 
    <audio src='https://0xxxxxxxx.execute-api.eu-west-1.amazonaws.com/alpha/files?file=https://xxx.s3-eu-west-1.amazonaws.com/file1.mp3&file=https://xxx.s3-eu-west-1.amazonaws.com/file2.mp3&file=https://xxx.s3-eu-west-1.amazonaws.com/file3.mp3'/>
</speak>

and I used this to confirm that S3 access works in the simulator:

<speak> 
    <audio src='https://s3.amazonaws.com/ask-soundlibrary/human/amzn_sfx_crowd_applause_05.mp3'/>
</speak>

Any ideas what is wrong? Is the issue with the http response from my lambda skill, or does something need to be enabled in api gateway? From my api gateway logs, it seems that the skill never tries to access the gateway.

Should I be using a different approach to fetch the mp3 for playback? Note, i want to use SSML as my audio is an effect and therefore shouldn't use the audioplayer (This is an Amazon requirement).

Upvotes: 1

Views: 290

Answers (2)

Ashish Yadav
Ashish Yadav

Reputation: 154

I might help you with this. The same problem happened to me. And after researching I resolved it. The problem here is "&" in you link which you provide in SSML. The solution you provided worked because there is no "&" now in your link. Too many parameters is not the problem.

I will suggest you to replace "&" with "&amp;"

In python -


url = 'https://0xxxxxxxx.execute-api.eu-west-1.amazonaws.com/alpha/files?file=https://xxx.s3-eu-west-1.amazonaws.com/file1.mp3&file=https://xxx.s3-eu-west-1.amazonaws.com/file2.mp3&file=https://xxx.s3-eu-west-1.amazonaws.com/file3.mp3'
url = url.replace("&","&amp;")


<speak> 
    "<audio src='" +  url + "'/>"
</speak>

I hope this helps you. Please let me know if doesn't work.

Upvotes: 1

Bailey
Bailey

Reputation: 167

Ok, I've worked it out myself.

It seems that SSML audio src doesn't like too many parameters in the URL call. I now just pass 1 parameter in the URL and use my lambda function to strip out the multiple filenames frome that single parameter.

https://0xxxxxxxx.execute-api.eu-west-1.amazonaws.com/alpha/files?file=/file1.mp3file=file2.mp3file=file3.mp3

Upvotes: 0

Related Questions