I am looking for an OSX editor software that can take an input file and find similar sounding files on my system based on its spectral content.
Any suggestions?
-
You would have to write it. It isnt very difficult task. have to read every file put the frequencies into a table, the more complex the anaylses the slower the search will be. then you have to compare that tune with every other in the table. It would take a good audio programmer for OSX about 10 days to complete a simple beta version. There is a program that can say the harmony infos of tunes for dj's i dont know it's name and its expensive, it does bpm and root key etc, not spectral though.– bandybabboonCommented Jan 12, 2016 at 0:57
Add a comment
|
4 Answers
The difficulty with the problem presented is that "spectral content" is not just a number you can compare to find degrees of proximity. You need to have complex analysis algorithms, comparing multiple parameters depending on the intended application, and define proximity criteria, normally based on statistical methods. This is true even for relatively "simple" sounds (musical instrument samples, or short natural sounds, for example), if the sound is too heterogeneous along time, the task is almost impossible and/or meaningless.
So in theory there could be (and perhaps there is, I don't know) a ready made tool to do, within certain constraints and field of application, the type of cataloging and searching described in the question. But what can be found more easily (normally within the realm of academic and scientific research) are resources and tools to allow to build such a tool to suit one's specific needs after (normally a lot) of experimentation and tuning.
One such tool is Essentia, a very powerful open-source C++ library for audio analysis and audio-based music information retrieval. This library has functions to evaluate dozens of different properties of sound spectra. A look at the Algorithms overview page, particularly the MIR (Musical Information Retrieval) functions is worth a look, to get an idea of the different audio properties that must potentially be considered.
This question at the Music.SE can also help to understand the approach to this kind of problem.
Things have come a long way in the last few years with regards to finding similar sounds in a sample library.
Some DJ software claims to offer key-matching algorithms to suggest songs so if you load your music library into Traktor et al, they may or may not do something akin to that. That is if you meant harmonic as in Harmony, rather than harmonics that create timbre
-
Thanks, I incorrectly said harmonic when I meant spectral and updated my question. There are similar functions in some synths to find similar sounding patches. I'd like to take a kick drum, and find similar files in my sample library based on it. Make sense?– RustyDevCommented Jun 11, 2015 at 18:53
-
It does make sense now but I don't know of a tool, sorry! It sounds like a sophisticated algorithm would be required...either the algorithm would return All other kick drums, or it would have to be tuned differently depending on whether a kick drum or a string or a horn or a kazoo was being input... Commented Jun 12, 2015 at 3:24
In order to do that, the software would have to read each and every audio file, I think this would be impractical, unless there was an initial scan, which took time to scan all audio files, created corresponding spectral information file/s to look back on whenever you did a search. I don't know of anything like that, most are tag reliant I think.
Synth patches would be different, they probably search for similar parameters(and therefore, sounds) which is much quicker(as they aren't audio files). Saying that, I'm sure they use tags as well.
