I can't get this to work. And the error messages are far from useful. First I tried on my own server: "I'm not sure what went wrong." I then tried hosting on S3, as they recommend: "There was a problem with the requested skill's response." I don't know if the SSL is wrong, or if it is not encoded properly. It looks to be encoded with MPEG-2, 48 kb/s, but FFMPEG has five million command line arguments. I'm not sure I'm using the right ones. Has anyone else got this feature to work?
Jo, Having similar problems. I am hosting my .mp3 files on S3, and getting the text response (with the speech tag) from a DynamoDB table. The echo speaks the "text" portion of the correct response, but when it gets to where the tag is, I also get a "There was a problem with the requested skill's response." I thought maybe I wasn't encoding correctly. I am using Audacity, saving as MP3 with the designated bitrate... Or, it May be an access error, where the Lambda service doesn't have access to the S3 file...but I have pulled my text responses from S3 in the past with success.
Yeah, I tried both audacity and ffmpeg. You can download the mp3 from the above links. When I do that and give it to an analysis tool it says it's MPEG-2 and 48 kbps. At first I had the kbps wrong, and Alexa did prompt on the screen that it was the wrong bitrate. Since it said that before, it [b]must[/b] be getting the file. (When I gave it http instead of https, it also complained, so that bit must be right too.) I have a whole bunch of miscellaneous sound files that I keep around. I was going to find somewhere/someway to upload them and leave them there as a public service for other Alexa developers. But that's moot if I can't get them to work.
I got it to work retrieving the same file mentioned by Galactoise, retrieved from the same place. But when I put that file in my S3 bucket, I am still getting the "There was a problem with the requested Skill response" error on the Echo. So it must be an access issue. I'm using cloudfront, maybe I haven't set that up properly, but I can access the files from a browser, so why not from lambda?
Yep, it was the mono/stereo thing. I now have a sound working when hosted on S3. I can only get the right output with Audacity. I can't work out the right arguments for FFMPEG. I try: [code] ffmpeg -i JET.wav -ac 2 -codec:a libmp3lame -b:a 48k JET.mp3 [/code] and it says [code] Stream #0:0: Audio: mp3 (libmp3lame), 22050 Hz, stereo, s16p, 48 kb/s [/code] But Alexa isn't happy. :-( If anyone works out the ffmpeg arguments, let me know. I've got dozens of little helpful sound files I'd like to put up for general use. Too many to do one-by-one in Audacity.
Hi all, We're very sorry for the confusion surrounding this. We've updated the knowledge base article here:
http://forums.developer.amazon.com/forums/thread.jspa?threadID=9685&tstart=0 The main thing to call out is that the sample rate is required to be 16000 Hz (though 24000 and 22050 Hz should work as well). The updated knowledge base article contains steps in using both ffmpeg, as well as Audacity for converting your existing MP3s over to this format. Again, we apologize for any inconvenience this may have caused you in troubleshooting this issue, and will continue to update any necessary documentation if any further issues are reported. Thanks!
@KJ - no, JET does not work. @Justin - Memo posted exactly what I was looking for. Unfortunately I have not been able to get it to consistently work. JET still doesn't work, but most other files do. I've gone ahead and done a batch convert and upload. Can some people try these out? They should (mostly) be ready to use in anyone's Alexa application. There are 300+ sounds there from a library I've kept on my disk. The web page lists them, and there is hover text which gives the audio tags. If a few people can confirm they work for them, I'll re-post under the Community Projects section.
Jo, I can report success with tags, when I save my audio files from audacity as described by Justin. Again, I am using cloudfront URLs pointing to mp3's I have saved on S3. I also got your "Trek - sl transporter" sound to work from my test skill, by pasting your link into my audio tag. Didn't try any others, but if you saved them all the same...
I saved them all the same. Wrote a batch processor for them. Yet some are still playable and others are not. I haven't dived into it to see why. Note: I wrote a skill, Xylene Playback, that I can queue up text via a web-app, and then just ask Alexa to play it back. This is very handy for bulk testing the output to fortune cookie skills. Turned out to be very handy for testing audio tags. I've considered submitting it and making it public, but I'd have to do account linking so each person would get their own playback stuff. Anyway. It's a pretty trivial thing. You might consider writing your own if you do this a lot.