Audio Playback about alphatab

[Danielku15]

Problem 1: The sample generation

The actual synthesis should be done in a WebWorker. The api would be something like this:

var synthWorker = new Worker('alphaSynthWorker.js');
synthWorker.addEventListener('message', function(e) {
    var data = e.data;
    switch(data.event)
    {
        case 'sfloaded': console.log('SoundFont loaded and ready'); break;
        case 'sfLoadFailed': console.log('SoundFont could not be loaded: ' + data.message); break;

        case 'midiloaded': console.log('MidiFile loaded and ready'); break;
        case 'midiLoadFailed': console.log('MidiFile could not be loaded: ' + data.message); break;

        case 'ready': console.log('Ready to play'); break;
        case 'synthCompleted': console.log('New buffer synthesized: ' + data.samples.length + " / " + data.position); break;
        case 'finished': console.log('Playback finished'); break;
    }
});

synthWorker.postMessage({'cmd': 'loadsf',  'url': 'mySoundFont.sf2'});
synthWorker.postMessage({'cmd': 'loadMidi',  'midi': Uint8Array});
synthWorker.postMessage({'cmd': 'seek', 'position': 8400});
synthWorker.postMessage({'cmd': 'synth'});

The Internet Explorer only supports WebWorkers in IE10 and Android since 4.4. (http://caniuse.com/#search=worker) . For the IE we could try to run alphaSynth completely in a Flash fallback solution, but Android <4.4 still has no solution.

[Danielku15]

Problem2: Sample Playback

The Web Audio API is quite new but already supported by Firefox, Chrome, Safari and Opera. http://caniuse.com/#feat=audio-api

A Flash fallback could be provided for the Internet Explorer.

On the fly synthesizing

The synthesizing and playing could be implemented like this http://stackoverflow.com/questions/6593738/audio-data-streaming-in-html5

alphaTab synthesizes the buffers, and new absolute timed buffers are played. Let's say we always fill 5 buffers and let them play. As soon one buffer finishes playing, we request a new filled buffer.

Synthesize all samples
The main problem in this case would be the memory. A 5 minutes song in raw audio samples will easily need >50MB. In addition it will take quite some time till the full song is synthesized.

But still this would be possible. And it would be easier to let a complete sample buffer play than stitch multiple small buffers together.

[Danielku15]

Flash Implementation

So far so good: A first version of alphaSynth is implemented and running within a hidden flash movie. I also have created the necessary extensions and jQuery plugins to integrate alphaSynth into alphaTab:

• alphaSynth loads a SoundFont2 file and a midi file generated from an alphaTab score.
• The position within the song (tick and milliseconds) is sent to JavaScript and position cursors are positioned on top of the canvas.

The API is currently a bit weak designed but for a very first version it's working. Next step is to generate the midi data correctly. It seems the generated midi file gets broken if there are multiple tracks. But this should be no problem to fix.