had another update,

wanted to let people know there's an unbiased comparison of key detection software going on at http://www.mixingonbeat.com/phpbb/v...p?p=21494#21494

so far 81 songs have been tested on songs where camelot has identified the key... see the link for details, but here is the summary so far:

rapid evolution 2:
43 correct (53%)
18 compatible (75%)
20 incorrect (25%)

mixmeister 5:
44 correct (54%)
12 compatible (69%)
25 incorrect (31%)

mixedinkey:
33 correct (41%)
24 compatible (70%)
24 incorrect (30%)

the percentage next to the compatible count indicates the degree of songs that are correct OR compatible with the actual key

things to note: this is using the latest version of rapid evolution which has a few improvements to the key detection (2.2.10). also, notice that mixedinkey is nowhere near the 95% accurate that it claims to be...

more songs will be tested, but i'm glad to see a side by side comparison actually done!!!

Good to see someone taking the time do do this. I guess nothing beats the ear then huh?

Hey guys I just wanted to say THANKS!
I am just now learning the ropes of hamonic mixing (I didn't know anything about keys until I read this thread), and being that I spin only vinyl I have actually quite a challenge since I cannot rely on software
I am using the Mixshare.com database to help me out, and I have a question about notation:
How comes certain tracks are indicated as A#m - 3A? Shouldn't 3A correspond to B-Flat minor (I am looking at the Camelot thingy).
For instance, Babe Instinct - Disco Babes from Outer Space, is indicated as A#m - 3A...
Am I missing something? Sorry if this is a retarded question but it's confusing the heck out of me...


Thanks again, keep it up

___________________
www.etn.fm
Imagination is more important than knowledge

quote:

Originally posted by qualia rapid evolution 2: 43 correct (53%) 18 compatible (75%) 20 incorrect (25%) mixmeister 5: 44 correct (54%) 12 compatible (69%) 25 incorrect (31%) mixedinkey: 33 correct (41%) 24 compatible (70%) 24 incorrect (30%)

um, so what youre saying is mixedinkey comes in LAST in the correct stakes after all of nennon's spam i wouldve expected better results.

let this be a lesson to you kiddies- learn to do it by ear!

___________________

quote:

Originally posted by SillyR How comes certain tracks are indicated as A#m - 3A? Shouldn't 3A correspond to B-Flat minor (I am looking at the Camelot thingy).

Same thing A# = Bb

quote:

Originally posted by skot_e Same thing A# = Bb

Thanks yes I figure it out now... reading through all the pages here

___________________
www.etn.fm
Imagination is more important than knowledge

quote:

Originally posted by pkcRAISTLIN um, so what youre saying is mixedinkey comes in LAST in the correct stakes after all of nennon's spam i wouldve expected better results.

yeah i know, i'm seeing tons of posts/advertisements stating mixed in key is so accurate and professional. it really has performed the worst.

during my testing i had an epiphany and was able to improve the key detection accuracy of rapid evolution significantly, it is now by far the best! here is the current summary with 91 songs tested:

rapid evolution 2:
58 correct (64%)
+19 compatible (85%)
14 incorrect (15%)

mixmeister 6:
52 correct (57%)
+14 compatible (73%)
25 incorrect (27%)

mixmeister 5:
51 correct (56%)
+12 compatible (69%)
28 incorrect (31%)

mixedinkey:
39 correct (43%)
+24 compatible (69%)
28 incorrect (31%)

i think rapid evolution had a bad reputation at first, because the key detection was pretty horrible when it first came out years ago. but it's come along away and hopefully these results will change that... you can view the details of my tests at: [[ LINK REMOVED ]]

I personally don't thrust this analyze by Cualia. One reason is that he is the sole developer of mixshare's rapid evolution and has been bashing "mixed in key" from the start. This just looks like propaganda to me. I want to see a objective and indepenteded analyz done by a "professonal". I would like to point out the only tru way of knowing the key is those done by a musician, and here no one can compete with camelots service.

On the other hand "mixed in key" is using a algorithm that is developed by a german company called zplane(www.zplane.de). Their technology is licensed by many well know companies like; ableton,alcatech(bpm studio), steinberg, native instruments, fruity loops etc etc.

Here is a little info about their key algorithm:
technology

***********
The input signal is analyzed about its tonal content (pitches) via a filterbank. The filterbank's mid-frequencies are constantly adapted due to the detected standard pitch to increase the detection accuracy. The resulting pitch distribution is analyzed with a sophisticated probability model to check for the most likely key acc. the detected pitch strengths. The detection accuracy is ca. 80 percent, compared to a random detection accuracy of only 4 percent. When classifying related keys as correctly detected, the accuration accuracy is well above 90 percent, making [tONaRT] a reliable and robust technology for automatic key detection. The standard pitch detection has an accuracy of app. 1-2Hz at 440Hz.
**********

And with new version coming up from "mixed in key" and a key detetion algorithm which is developed by a large and respected company I believe "mixed in key" is the best software choice. And no, I am not getting paid to say this :-)





kind regards
djprince, norway
www.djprince.no
Online since 1996 and key/bpm database/community since 1999
PS! site is currently under updating

Last edited by djprince on May-18-2006 at 09:38

i use a program called \'ears\'. it has a 95% accuracy rate.

___________________

i dunno pkc, I'm not so sure that my version of that prog is as accurate. I've got nothing to check it against tho. Maybe it just needs to be calibrated over time.

quote:

Originally posted by pkcRAISTLIN i use a program called 'ears'. it has a 95% accuracy rate.

The way to go!


DJPrince,

Thanks for the heads up on Qualia, but he has been around long before this thread started and is a propper TA who just happened to become very passionate about Harmonic Mixing. He has also contributed a lot of useful information in various threads and has only ever really been out to try and help the online community much like yourself.

I do however totally agree with you on the fact that you can't beat getting your tracks keyed by ear. I produce my own tracks etc and do this all myself.

The problem with Camelot is that for the dance music industry they are so chronically out of date, or simply do not have the tracks you are looking for partly due to the nature of dance music and therefor I would never recomend them to any trance DJ in terms of a source for keyed music.

Cheers
Nem

___________________
https://www.mixcloud.com/Calvin_Karass/

My ears are definitely not trained enough to detect the keys consistently so right now I run my tracks through Mixmeister, then verify the result with the detector in Rapid Evolution while having the piano open. It seems like RE catches the keys that MM gets wrong most of the time, but sometimes you have to run the detect function a couple times just to be sure while you're verifying it on the piano.

It works for me right now and you can never be too sure.

Hopefully in time my ears will catch up : )

___________________

quote:

Originally posted by Dirk W. People in Dallas aren't *ssholes. Just look at me, I live in Dallas.

Meat is Murder...tasty, tasty murder.