REKORDBOX 2.01 Now with Key Detection!

[and_leite]

I've set up an action in mp3tag that automatically converts all of the keys that get written by rapid evolution to camelot keys. Then I can just load my tracks into mp3tag, start the action and wait for them all to be converted!

Man, I never thought of using an action in Mp3tag. This makes the process so much easier. Thanks for the tip.

[Stereo:Type]

I had a look at this out of curiosity as to how accurate the key detection was. I actually thought it worked better than MIK. I analysed about 30 tracks and it got 3 or 4 wrong. The difference i found though was that where as mixed in key will be wrong to a relative degree (ie: it might identify a 5A track as a 6A because there could be a predominant part of the music with a G as opposed to a C)
the ones that were incorrect in Rekordbox were just flat out wrong.

I'm not sure why so many people rave about mixed in key. I personally think their system is a bit of a pisstake. It's such an inconvenience for users to have to be online to have their tracks analysed. What if you're on a plane on the way to a gig or in the middle of Backwater, Nowhere without internet access doing some prep work? They're so precious about other people getting at their 'Special Algorithm' for detection. They would probably have more reason to be so protective if mixed in key didn't give incorrect results half the time (albeit to a relative degree). Added to which their 'revolutionary' Camelot wheel is just a theoretic principal which has been around for eons (though credit to them, looking down a list of numbers is easier at speed than a list of letters with Flats and Sharps)

Bottom line, i think Rekordbox does a better overall job if you want key analysis. If i didn't do it by ear, i would want a program that got it totally wrong 10% of the time over a program which made practical use difficult as well as constantly half assed results...

[wys01]

No key detection algorithm is 100% accurate (even different people will key the same track differently) so you can pick a small selection of tracks and it will appear that one is much better than the other. I don't think trying 30 tracks is really enough of a trial to say if MIK is better or worse than Rekordbox.

Also you seem to say that MIK got keys wrong to a relative degree whereas Rekordbox got it totally wrong, do you really mean that RekordBox is better than MIK? Being slightly out is surely better than being completely out, no?

[Stereo:Type]

I dunno, i would say that 30 is enough to give you a representation of it's accuracy. If i did 300 you would only expect to see that number multiplied ten fold...

I guess there is some bias in my opinion because i feel like MIK takes people for a bit of a ride by exaggerating the capability of their software, and also by their practise of forcing users to be online to use their software. I know from my experience that were I a user of it, I could think of a number of times where that would have screwed me over...

In this particular case, I think being wrong to a relative degree and being totally inaccurate hold different pitfalls.
What's messed up about the way that MIK is wrong is that if say, you have something in 4A and it labels it 5A, then you are never going to pull out a truly harmonically balanced mix unless you go 5A->4A. So every time you mix 5A->5A with that track, then you're pulling out a mix which could still be pretty tenuous. Just because something is a 5th up doesn't mean the mix wouldn't sound ropey as hell, it just means that its the best chance you have of modulating from one key to another.
Added to that if you try to go 5A->6A when in reality it's 4A->6A then that is most likely going to sound like a total train wreck. Now lets say you have the 5A (really a 4A) and mix it with another incorrectly labelled track, this time a 6A which is really a 7A. Now, the user thinks they are going from 5A->6A, but in reality they are moving 4A->7A.

Long story short, if you have inconsistencies in the labeling even to this relative degree, if those are all over your library, then the system pretty much falls apart. At the very least, it's hard to put your trust in it...

When something is flat out wrong it's easy to tell. If i was using key detection software, I would rather the mistakes be obvious when they are there, but the rest of the time be able to trust as much as is possible that the software is giving you the correct answer. That's why i think that Rekordbox will probably steer people who want this kind of function in a better direction, rather than MIK which uses an over inflated opinion of its-self to give DJ's a pretty false sense of confidence in the matter.

Also, if two people come up with two different answers as to the key of a track, then either one of them is wrong, both of them are wrong or the track changes key somewhere...

[wys01]

I think what you're describing is the trade-off that you make when deciding to either key your tracks yourself or let some other program do it for you. Doing it yourself is going to give you the most accuracy, but will likely take much more of your time than running them through a key detection program. If you can cope with less than perfect accuracy of your keys then it's a very useful tool and will save you a lot of time.

I agree the blurb on the MIK website about "virtually guarantees" and "100% harmonic" can give the impression that it hardly ever gets the key wrong, but it's not saying anything technically wrong.

The key of a track is subjective, two people can disagree about the key and both be right. Most of time the key is pretty clear cut, but sometimes it can come down to a choice of two keys and some people perceive the track to be in one key rather than the other, without there being an actual key change in the track.

[fullenglishpint]

I think you're really putting far too much emphasis on keys. If it doesn't fit, you can hear that in your headphones. As you said, when something is flat out wrong it's easy to tell. If it's not flat out wrong, you can make the mix work. Even if it is flat out wrong, you might be able to use a sparse section or loop without melody.

Also, a given algorithm might be much more accurate with certain genres or styles. 30 songs isn't enough to accurately gauge that.

[Morgraw]

quick question: how can I update my traktor library after I analyzed the track in rekordbox? the Key doesn't appear!
maybe the tracks must be unlocked?

[canzil]

quick question: how can I update my traktor library after I analyzed the track in rekordbox? the Key doesn't appear!
maybe the tracks must be unlocked?

Try select tracks and check consistance.

[fullenglishpint]

I'm not sure if rekordbox stores key data in metadata or in its own separate settings file.