Details about as3 Sound.extract() format? - actionscript-3

Question

Using the sound object .extract() method, it's possible to retrieve the uncompressed sound data as a ByteArray. What is the format of the extracted sound data? Custom internal format, or established format? Endian type?
I would like to be able to load sounddata directly into a byteArray and feed sound playback from this data (bypassing step 1 below).
Here are my experiments so far:
Step 1. Here I load a sound, extract the data content and save it to disc:
private var sound:Sound;
private var byteArray:ByteArray;
// loading a sound...
private function init():void {
sound = new Sound();
sound.addEventListener(Event.COMPLETE, onComplete);
sound.load(new URLRequest('short.mp3'));
}
// saving it to disc...
private function onComplete(e:Event):void {
byteArray = new ByteArray();
sound.extract(byteArray, 100000000);
byteArray.position = 0;
var file:FileReference = new FileReference();
file.save(byteArray, 'short.data');
}
Step 2. Here I load the data from disc and feed the playback of a new sound object with this data:
private var sound:Sound;
private var byteArray:ByteArray;
// loading a sound...
private function init():void {
file = new FileReference();
file.addEventListener(Event.COMPLETE, onLoaded);
file.addEventListener(Event.SELECT, onSelected);
file.browse();
}
private function onSelected(e:Event):void {
file.load();
}
private function onLoaded(e:Event):void {
trace('loaded');
trace(file.data.length);
byteArray = file.data;
sound2 = new Sound();
sound2.addEventListener(SampleDataEvent.SAMPLE_DATA, onSampleData);
sound2.play();
}
private function onSampleData(e:SampleDataEvent):void {
trace(e.position);
for (var i:int = 0; i < 2048; i++) {
var left:Number = byteArray.readFloat();
var right:Number = byteArray.readFloat();
e.data.writeFloat(left * 0.2);
e.data.writeFloat(right * 0.2);
}
}
I would like to be able of accomplishing this second step (loading data from disc and feeding a sound object playback) by using an external converter utility (sox or something).
/ Jonas

Answer 1

It should be standard 32bit floats, one per channel per sample. So I would try -e floating-point for sox, I haven't tried though but I would do it with a certain level of confidence... :) Floats are endian-independent, I believe...