Connexion rapide:  

Forum: General Discussion

Sujet Virtual DJ 8 BPM issue - Page: 4

Cette partie de ce topic est ancien et peut contenir des informations obselètes ou incorrectes

SerinePRO InfinityMember since 2013
I regret even bringing that issue up now. If you download some of those tracks (yep, I've even provided download links) you can very clearly see that the BPMs are generally integers and xxx.000 is the correct BPM which lines up best with beats at the beginning, middle and end of the song.

I understand that this is not the case for many tracks and varies across genres but for these specific tracks and most of the tracks in my library, it is. I understand that it is especially not the case for music played by humans. I understand that it is uncommon for older digitally produced music. But for these tracks, it is the case.


Again I'm not really interested in what "can" happen or what "real DJ's" can or "should" do. I am interested in the fact that in the vast majority of cases I encounter the BPM analyzer is not working at all. This means that the automix feature, which I use daily, does not work and it means that novice DJs interested in using VDJ may decide to use a different piece of software.
 

Posté Wed 18 Jan 17 @ 2:41 pm
Serine, I think it is a very good subject to bring up and has come up before. People will argue this and that but it can help to give a better understanding of the whole concept. Maybe not for some who have been doing it for many years, but definitely helpful for some.
 

Posté Wed 18 Jan 17 @ 2:48 pm
SerinePRO InfinityMember since 2013
I agree, but maybe it would be good to do that in a thread where the examples don't disprove your point, or somewhere that they won't need to read through 3 pages of technical discussion? Or make it clear that it is being offered as additional information rather than a rebuttal?

Because from here it feels like I'm trying to have a discussion about the BPM analyser and have gone to some trouble to provide a variety of mp3 and jpg examples, and people are ignoring it so that they can remind me that they are better DJs than I am and that they know more about music, despite me repeatedly having humbly agreed that I am a s**t DJ and I have a limited knowledge of music.
 

Posté Wed 18 Jan 17 @ 2:53 pm
I am in the process of deleting all the duplicates out of library. Some tracks I have 3 or 4 times as they have been ripped from different albums. Most of the time though, they all have a different BPM, even though they are the same track. They generally don't differ by anything more than 0.5, but it is interesting that they differ at all. I can only assume that some have been remastered, that the endings have been shortened or faded or something. I have no idea in all honesty. But whenever i see it, I do find it curious.
 

Posté Wed 18 Jan 17 @ 4:09 pm
AdionPRO InfinityCTOMember since 2006
The 'first beat" poi is only used for the beat grid and phrase meter, so while it is usually easiest when manually setting it to set it at the actual first beat, for vdj it's perfect as long as it's a multiple of 16 away from the first beat.

For automix the automix poi are used (see show all in the poi editor). Which ones depends on the selected automix mode.

The detection in vdj ranges from 50 to 200, but if a bpm is already available, the range (50 to 100 or 100 to 200) is kept from the previously known bpm.
 

Posté Wed 18 Jan 17 @ 4:19 pm
SerinePRO InfinityMember since 2013
But automix uses the first beat marker to set the phrase, right? So the automix point can be set to the second beat in a phrase but the software is smart enough to still make that sound good. I'm just doing it now and it works fine, with the phrases perfectly aligned no matter where the automix POI is, but sounding awful if the first beat POI is incorrect.
 

Posté Wed 18 Jan 17 @ 4:51 pm
AdionPRO InfinityCTOMember since 2006
Correct, the grid is still used to match the songs
 

Posté Wed 18 Jan 17 @ 5:08 pm
The fact is, no automatic mixing software is goning to do a perfect job. That's why DJs are still in business!

There are lots of apps for example that run on smartphones and tablets, which take your track library and "mix" it together for you. I've got a few myself, just for fun. They've all got shortcomings.

The best one I've heard was developed by a guy (Ian Chamings) who was on Dragon's Den a few years ago. He got backing from two of the dragons / investors. Initially his idea was to sell custom mix CDs to the public, but it's ended up as a service for fitness pros who want mixes for workouts. His algorithm was pretty smart. Not available in DJ software though. :-)

Google his name for details
 

Posté Wed 18 Jan 17 @ 6:45 pm
SerinePRO InfinityMember since 2013
I don't need it to be perfect, I just want it to work. Again, I think you are missing the point.

If the "first beat" is being detected as silence exactly 1 beat before the first beat, that seems fixable to me. Instead of pretending it's not a problem and telling me, again, for the nth time in this thread, that DJs make software obsolete, I am trying to provide constructive feedback and examples so that the development team can understand what the problem is, who it affects and whether they can or should fix it. The answer could be that they don't want to, or that they can't and that's fine. As I have said before, I think this is a great piece of software and I bought it on the understanding that this feature wasn't supposed to work.


Apologies if you just mean to add that information as a point of interest, but it comes across as you trying to correct me and tell me that this feature should not be improved.
 

Posté Wed 18 Jan 17 @ 6:58 pm
What I'm saying is that automated mixing is not something that's particularly good in any software (that I've encountered), not just VDJ - so don't get your hopes up.

If it was an easy thing to implement then all the software out there would do an amazing job blending tracks together, but it doesn't.
 

Posté Wed 18 Jan 17 @ 7:50 pm
I see Serine's point here - i have noticed a significan number of tracks in my library that need manual correction of the beatgrid phase by plus or minus one beat. So many that you would suspect an algorithmic tweak could be applied.

The beat detection is pretty good - but in my experience the phase for some reason is more often than not wrong (but not always).

Some are tricky obviously where you have a song that slips in some funky bridge ,verse, chorus or break lengths. There are good reasons why it is not perfect - but maybe the question should be - is there a way to make it more correct more often?
 

Posté Wed 18 Jan 17 @ 9:16 pm
 

Posté Wed 18 Jan 17 @ 11:03 pm
SerinePRO InfinityMember since 2013
Can we not use this thread to argue about the easily solved issue of half/double time detection please?
 

Posté Wed 18 Jan 17 @ 11:09 pm
The range does not only effect halving and doubling. It can on occasion effect an incorrect BPM detection. Unless you take this on as a whole and the sum of it's parts it will never be correct. So you say you have one issue and nothing else matters. Trying to correct that one issue can lead to other problems. This is a complex subject. It is so complex that the state of the art in beat detection has to do with neural networks which is complex in itself. Neural networks model the way the human brain thinks and it was a break through in understanding many years ago. Most of the time you do not need to go to this extent to solve some computer problem. Think about it. Why does a BPM detection algorithm need a range in the first place? Humans don't need it. It is because the BPM detection code is not that smart yet. It is unknown how it will be improved but there is no shortage of research on it.

I think they are trying to address your issue the best they can. At any rate most people can tap their feet and rub their belly at the same time :) meaning it is easy for most to talk about different parts of an issue without distraction.
 

Posté Thu 19 Jan 17 @ 3:55 am
SerinePRO InfinityMember since 2013
I just bought Hospital Records: We are 21.

Listening through on automix reveals a good 50% of the tracks with completely incorrect grid detections.

Mat Zo - MAD (mp3) is a good example of the incorrect BPM detection with 0.008% error actually being quite audibly out of sync by the end of the track.

If I Move To Mars (Urbandawn Remix) (mp3) is a good example of one where the first beat is incorrectly detected about 10% of the way through a bar with a really obvious first beat shortly before it and no beat to speak of at the point it detects.
 

Posté Mon 06 Feb 17 @ 6:50 pm
I've just encountered a problem with VDJ 8 (3710) analyzing the BPM of Be My Lady by Kool & The Gang.

It's supposed to be around 125 BPM (and Mixed In Key gets it right) but VDJ's first attempt tells me it's 83.1 BPM.

If I switch to the new experimental analyzer it says 62.3 (which I guess is half the correct tempo at least).

Another example - Cosmic Lust by Mass Production. Again VDJ says it's 84.2 when it's actually around 128.

Anton Powers & Pixie Lott - Baby (VDJ says 81.3 and it should be 122).

[later]
Now this is strange......the above results were from my laptop (Windows 10 64 bit). On my older desktop PC (Windows 7 32 bit), VDJ gets them all right first time. Why should that happen?
 

Posté Mon 01 May 17 @ 2:50 pm
VDJ RonPRO InfinityMember since 2010
I think that the guys pointing out the grid being off by a few 1/1000ths are forgetting that a beat can have swing applied to it.

That's a good point Locodog.

I would not have known what you were talking about except that I've been looking at the Traktor step sequencer which has numerical values of swing varying from 1-80. The idea being to randomize the music making it sound more human, and less electronic. This of course makes it more difficult to mix.
 

Posté Mon 01 May 17 @ 4:02 pm
bigron1 wrote :
This of course makes it more difficult to mix.


No no - swing won't change the overall BPM of the track. It's not about making the whole BPM vary.

Given a 120 BPM beat with every single sound exactly "on grid", swing will just very slightly shift some of the elements a tiny amount before or after the "grid".

Usually not the kick drum. Swing will usually apply to the snare hits, hi hat and other percussion elements (bongos etc).



 

Posté Mon 01 May 17 @ 4:53 pm
VDJ RonPRO InfinityMember since 2010
The overall bpm is not what's important..it's the local bpm, and the beatgrid over the regions being mixed which is critical.

Using the step sequencer I can construct a 4 beat bar using a variety of samples such as kick drums, hi hats, sticks, and so on.. placing these samples originally on the beat or even on the 1/4 beat but then using swing I can move each individual sample off beat There seems to be room for quite significant off beat displacement. Is it not the case that songs might be constructed in the same way?

Snare hits are loud and significant hm..
Interesting Usually not the kick drum.
I've been moving the kick drum around lol.
My step sequencer usage is a little chaotic at the moment ..more lol.



 

Posté Mon 01 May 17 @ 5:09 pm
Here's a page with some videos about adding swing (groove) and generally humanising things with Ableton Live.

https://www.ableton.com/en/blog/get-swing-drum-programming-tips/

One wouldn't normally do this with a track that's already played by humans (i.e. pre drum machine). Apply it to a really stiff robotic track though and it can change the feel - but don't overdo it. The groove profiles used by DAWs are usually very subtle.

Probably the most famous drum pattern in the world - Funky Drummer, played by Clyde Stubblefield - is so popular because it has feel/groove/swing. It's mentioned in this video https://www.youtube.com/watch?v=Z3xSXc1vy5I

In this video you can hear the "flam" sound created by the two drummers being slightly out of sync with each other (they're human). https://www.youtube.com/watch?v=rE65eaI9JtU
 

Posté Mon 01 May 17 @ 5:30 pm
79%