Recent Posts

Pages: 1 2 [3] 4 5 ... 10
Plugin Development / Vamp plugin SDK v2.6 now available!
« Last post by cannam on June 16, 2015, 14:11:42 »
Version 2.6 of the Vamp plugin SDK is now available.

Vamp is a plugin API for audio analysis and feature extraction plugins written in C or C++.  Its SDK features an easy-to-use set of C++ classes for plugin and host developers, a reference host implementation, example plugins, and documentation.  It is supported across Linux, OS/X, and Windows.

A documentation guide to writing plugins using the Vamp SDK can be found at

Version 2.6 is a bugfix and minor enhancement release. For more details, see the changelog at

Thank you Chris! That was very thoughtful and will have a look into it. Yet I will have to decide how much time I will preserve for testing the tools as such vs actually starting to compare the different interpretations.

The text link is great! Thanks!
Host Forum: Sonic Visualiser / Re: MP3 or WAV different result?
« Last post by Apfelmatsch on April 29, 2015, 14:21:32 »
Thank you for answering!

OK, the pdf you linked does sugguest (as you somewhat gave a hint to) that a minimum of aprox. 160kbps is needed to not affect the research results.

I am still a bit insecure, though. There seems to be a special case: What is described here, is that different from the loudness war phenomenon (which will not play a role here)? If the studio master has already been made very loud, it will get problematic after compression. So let's say from the tape, they made an audio CD setting the loudest point as high as possible or even upping the most silent points of the tape. I guess when the MP3 seller / the label then compressed that to mp3 you get a similar phenomenon like in loudness war (I guess that was what you were referencing to radio broadcast situations etc.).

Well, ok. I can never know, what the studio guy did to the tape. When the conductor conducted from pianossimo to fortissimo and the studio guy just equalized that, I can't do anything about it. I probably won't even know, if he did change the conductors intents.

Apart from that is the loudness war stuff been done to classical music, too? Maybe in a false attempt to make it more impressive to the listener?

Another thought: according to the pdf, in my tests, could I go so far and even compare 256kbps mp3s (as I guess that is the most likely to be found) or 320kbps mp3s with FLAC/ALAC and wav/aiff? Like conductor 1 as MP3 and conductor 2 as wav (which would be derived from FLAC) and compare them?

Maybe I should even compare mp3s of the same conductor and same recording against each other, maybe I find that different mp3 sellers used different quality compressors (given that all use 256mbps).

Since I have a Mac and Sonicvisualizer does only take wav I will also have to look out for something that can convert FLAC (PC) to wav and ALAC (Mac) to wav and maybe even aiff (Mac) to wav (PC).
Hi -- this is a good general question but it is also a rather difficult high-level question, and one might need quite a bit of experience in real musicological work to be able to answer.

While Sonic Visualiser and the available plugins provide some tools for breaking down music in interpretable ways, it is still very hard to work backwards from a recording to details of things like instrumental balance or local dynamic or tempo variation in a classical work, not least because most methods are primarily designed for pop/rock music. (Even local tempo estimation and "beat tracking" are essentially unsolved problems for classical music.)

One thing that is quite valuable is to load multiple recordings into Sonic Visualiser and then time-align them -- for this you need the MATCH plugin installed. This can give you some information about the conductor's decisions in terms of both overall and local tempo. There is a video showing this at (Audio alignment using Sonic Visualiser and MATCH).

The videos from Musicology for the Masses may also be of interest, as well as "A musicologist's guide to Sonic Visualiser" (, but (as you have probably seen already) these primarily focus on Sonic Visualiser's built-in visualisation tools rather than higher-level analysis using Vamp plugins.

It's quite possible you may get interesting results from plugins that analyse tonality (possibly even the "Key Strength Plot" from the very basic key estimator in the QM Vamp Plugins set) or loudness (there are various options for analysing amplitude -- we have a proper auditory-model loudness plugin in the works but it isn't quite published yet).

Host Forum: Sonic Visualiser / Re: MP3 or WAV different result?
« Last post by cannam on April 24, 2015, 14:17:54 »
Hi there -- please don't worry about asking stupid questions, this is a very reasonable one.

Generally speaking the compression used for mp3s does not make a difference to dynamics and the like -- it has nothing in common with the dynamic range compression typically used for e.g. radio broadcast. A good-quality mp3 should give you essentially the same results as the wav version for reasonably robust audio features, although you may get different results for very low-level features (such as those sensitive to inaudibly high frequencies appearing in the signal).

There have been various publications on this subject, for example here:

Do be alert, though, to the fact that there may be padding differences between different encodings or releases of the same piece (i.e. different amounts of silence at end or beginning).

What Plugins are needed to compare parameters of different musical interpretations (recordings) of classic music? In this case a Symphony conducted by different conductors.

do I need any of the plugins to have a complete test set for lets say "conservative" use? (compare what the conductor let the orchestra let sound/play/do different. Of course I know that the guy at the mixer will have an influence, too to some degree, but... anyway.)

It this is a stupid question, please don't ignore, but tell me, that it is stupid, so I don't wait for answers any longer.
Host Forum: Sonic Visualiser / Re: MP3 or WAV different result?
« Last post by Apfelmatsch on April 24, 2015, 14:02:56 »
Does no one have an answer or did I ask a stupid question?
Host Forum: Sonic Visualiser / MP3 or WAV different result?
« Last post by Apfelmatsch on April 21, 2015, 21:50:40 »

I'd like to know, if it will have an effect on the test results, if I use MP3s or WAV (i.e. AIFF = CD-audio).

I imagine, since MP3 is more compressed, the used codec might have e.g. compressed the loudness of the music, too.

I have to compare interpretations of a neo-romantic/modern symphony and while getting MP3s of all interpretations would be the easiest and cheapest way, I fear I will miss something of the original that got compressed out of a studio file, while I think the wav/Aiff/CD-Audio will have more of the orifinal studio/live-recording.
Also I guess, I should not mix MP3 and WAV interpretations, right?
Also, I guess, I should not use some MP3s from Spotify and compare to some MP3s converted from CD-audio with iTunes or mepgstreamclip or different MP3-portals/streaming-services, right?

Well, of course, I could just test it and make it part of the text I am working on.

Thank you.
Getting and Using Vamp Plugins / libaubio.dylib
« Last post by Shapour on March 21, 2015, 22:52:27 »
As I am trying to install/use plugins, the program does not recognize them. Here is out put of the program.
Am I supposed to have the libaubio.dylib already under my /usr/local/lib? Where do I get it from?
Am I supposed to do something before simply copying the three plugin files into the /Library/audio/Plug-Ins/Vamp?
WARNING: FeatureExtractionPluginFactory::getPluginIdentifiers: Failed to load library /Library/Audio/Plug-Ins/Vamp/vamp-aubio.dylib: dlopen(/Library/Audio/Plug-Ins/Vamp/vamp-aubio.dylib, 5): Library not loaded: /usr/local/lib/libaubio.dylib
  Referenced from: /Library/Audio/Plug-Ins/Vamp/vamp-aubio.dylib
  Reason: image not found

Duplicate statement encountered on import in ImportFailOnDuplicates mode [with triple <( <> <> "Compute energy in each 40 mel-frequency bands" )>]
PluginRDFIndexer::pullURL: Document at duplicates triples found in earlier loaded document -- skipping it
ERROR: PluginRDFDescription::indexURL: No vamp:identifier for output <<>>
ERROR: PluginRDFDescription::indexURL: No vamp:identifier for output <<>>
Host Forum: Sonic Visualiser / Re: File size/length limit?
« Last post by cannam on March 03, 2015, 15:37:15 »
Hi Frazer -- I don't know about alternatives, but I can confirm that (sadly) SV can't currently handle this. Although SV is designed to handle files and analysis plots that don't fit in memory, there are still 32-bit length limits for some calculations internally.

This is one of the things that I'm working on in the background at the moment -- we've seen a few requests for a true 64-bit build for Windows (the current one is 32-bit only on that platform) but there doesn't seem much point while we aren't properly taking advantage of the architecture -- especially since all the current plugin builds are also 32-bit only and a 64-bit host can't currently run 32-bit plugins.

Pages: 1 2 [3] 4 5 ... 10