Re: Web Audio Processing: Use Cases and Requirements

Also, AudioBufferSourceNode and AudioBuffer together provide the ability to emit sound based on an arbitrary array of sound samples, which is more general than an Oscillator. The Oscillator is very handy for optimizing the particular case of common periodic waveforms.

…Joe

On Sep 22, 2012, at 11:15 AM, Chris Wilson <cwilso@google.com> wrote:

> Hello David,
> You are quite correct in that the use cases document does not have many scenarios for synthesis of audio; it does mention synthesis prominently, but the only explicit synthesis use case mentioned is synthesizing a metronome sound.  We should make that more explicit.
> 
> That said; the Web Audio API does have quite a powerful Oscillator node to synthesize sounds, and many of the features of the API have been designed to enable common synthesis needs.  In fact, I put together a synthesizer that uses Web Audio to build a standard analog-era synthesizer - http://webaudiodemos.appspot.com/midi-synth/index.html - with no sound samples involved (actually, not entirely true - I do use an impulse response sample for the reverb).  The AudioParam scheduling mechanism enables a lot of powerful envelope controls.
> 
> Note, by the way, that Ray's blog post was talking specifically about HTML5 <audio> - that is, the HTML audio element - not the Web Audio API, which was still very new when he wrote that post.  A number of the samples Chris Rogers has written (at http://chromium.googlecode.com/svn/trunk/samples/audio/) are sound synthesis demos - e.g., all the oscillator-* files, wave-table-synth.html, and tone-editor.
> 
> -Chris
> 
> On Sat, Sep 22, 2012 at 6:51 AM, David Dailey <ddailey@zoominternet.net> wrote:
> I am pleased to see the work on this topic  [1].
> 
>  
> 
> The use cases seem to lack something that, in my mind, is rather fundamental: the ability to create sounds ex nihilo.  In the 1980’s Mac users had access to a pretty little program called SoundEdit [2]  that allowed one, using SVG-like shapes (though I don’t recall that we called it SVG back then) to create waveforms that were then converted to simple sounds. A sine wave of a particular frequency might correspond to a pure tone. Waveforms could be combined to create timbre, so that voices could be created. Throughout the document, I see lots of references to using pre-recorded sounds, stored as little “auditory bitmaps” somewhere, but nowhere that a composer could construct the primitive sounds herself.
> 
>  
> 
> I think I might not be the only person interested in such.  Ray Cromwell’s blog [3], mentioned at [4], points out an inability of HTML5 audio: “you cannot synthesize sound on the fly.”
> 
>  
> 
> Perhaps this is at the core of people’s thinking already and that it has, accordingly, been so obvious as to elude mention. Perhaps I’ve missed it in my perusal of the use cases (apologies, if so – it would not be the first time I’ve misread such things).  In my own shallow and brief experimentations with computer generated music over the past 4 decades, the generation of primitive sounds would seem to be important to the group’s efforts.
> 
>  
> 
> I would suggest that something like InkML with SMIL and a <path>-like element that has PostScript-like loops, recursions, reversals, transpositions and the like would go a long way once the composer can create (or borrow) a set of notes and voices.
> 
>  
> 
> Regards
> 
> David
> 
>  
> 
> [1]  https://dvcs.w3.org/hg/audio/raw-file/tip/reqs/Overview.html#music-creation-environment-with-sampled-instruments
> 
> [2] http://en.wikipedia.org/wiki/SoundEdit
> 
> [3] http://cromwellian.blogspot.com/2011/05/ive-been-having-twitter-back-and-forth.html
> 
> [4] http://lists.w3.org/Archives/Public/public-audio/2011AprJun/0041.html
> 
> 

... .  .    .       Joe

Joe Berkovitz
President

Noteflight LLC
Boston, Mass.
phone: +1 978 314 6271
www.noteflight.com