GSOC Analyzer Support in Amarok and possibly Phonon

Ricard Marxer Piñón email at
Tue Mar 3 16:23:05 GMT 2009

On Tue, Mar 3, 2009 at 4:00 PM, Ian Monroe <ian.monroe at> wrote:

> On Tue, Mar 3, 2009 at 3:29 AM, Ricard Marxer Piñón
> <email at> wrote:
> >
> >
> > - Show quoted text -
> > On Sat, Feb 28, 2009 at 6:12 PM, Michael Pyne <mpyne at>
> wrote:
> >>
> >> On Saturday 28 February 2009, Ricard Marxer Piñón wrote:
> >> > The main idea is to make audio visualizations (using Phonon to access
> >> > audio
> >> > data) for Amarok 2 similarly as what was available in Amarok 1.x.
>  Most
> >> > visualizations there were spectral based and we would then need some
> >> > processing of each audio frame to be able to this.
> >> >
> >> > Option A means doing the processing outside of the Phonon backends
> >> > (personally I would choose FFTW and Eigen to do this processing).
> >> > <snip>
> >>
> >> I like Option A since it doesn't add much in the way of external
> >> dependencies
> >> to Phonon.  Option B seems to do better at making the functionality to
> all
> >> Phonon users though.  Maybe just have a very thin plugin interface that
> >> Phonon
> >> could use to list visualization plugins with the application being
> >> responsible
> >> to host the plugin?
> >
> > Hi, and thanks for the reply.
> >
> > I agree that to let Phonon be as lightweight as possible these sort of
> > plugins should be hosted by the application.
> Though the basic analyzer, intensity of sound at various frequencies
> should be pretty trivial. You don't need FFT for that.

Hm... yes, it could be done with a time domain filterbank but I think this
would be more cumbersome (depending on the frequency resolution it could
even be more computationally expensive) and FFT is not that heavy.  But I
can easily make both and check, that wouldn't be a problem (I actually
already have some code for each of these, will clean it up and see).

Anyway, I think an AudioSpectrumOutput class could be of general interest
and very useful for visualization purposes.  Furthermore with only a few
easy to understand parameters (number of bands (frequencies), intensity
scale (log vs linear)) it could be very useful for all the people planning
to do the content base audio analysis (I'm thinkning of all the GSOC
proposals about intelligent playlists in Amarok).

We can later decide where this class would fit depending on it's usefulness,
weight, computational cost and of course implementation quality.

I will write up a draft proposal this week which will include:
 - AudioSpectrumOutput class which could at some point end up in Phonon
 - A visualization for amarok which will use that class
 - And a plasma applet wrapping ProjectM which will use AudioDataOutput
class from Phonon::Experimental

> > By the way, am I on the wrong list? Ian told me about a list called
> > phonon-devel but I haven't found that one and the Phonon site directed me
> > here.
> I really did say phonon-backends, not phonon-devel, I just checked. ;)

doh! sorry for that.

> Ian
> _______________________________________________
> kde-multimedia mailing list
> kde-multimedia at

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>
-------------- next part --------------
kde-multimedia mailing list
kde-multimedia at

More information about the kde-multimedia mailing list