Beat grit does not match the song

Hello everyone,

I have a problem that I cannot understand.

In engine DJ that runs on a MacPro I have the problem that the beat grit does not match the song for many of my 18000 songs.

Accordingly also not the BPM and the KEY. All songs have been checked with the auto analysis and it does not affect all songs. But quite a few.

It is also not about variable BPMs in the songs that are not analyzed correctly. Everything that is not dance or electronic has problems. I’ve also had this in Serato, but not as extreme.

I can’t get any further here. Thanks for your help

1 Like

Moin @Rene_Krause,

it’s a never ending story. In the link below you can start at the roots’ of the discussion.

https://community.enginedj.com/t/flexible-beatgrid-ability-to-set-a-manual-downbeat/36984/1

Since the first post DENON amended a lot, but it’s not a finished story yet.

A lot of tracks need to be improved by hand, it takes substantial experience and it’s hard work to improve the tracks, especially Disco Tracks from beginning of Disco. A challenge is to adjust a 6/4 rhytms to 4/4 rhytms.

The tools DENON offers for this improvement by hand are in ENGINE DJ (Grid Edit Controls item 9) and in Engine OS (35. Edit Grid: Press and hold this button to enable beat grid editing).

Good luck and brgds from HAM to HAJ

BeatMaster

2 Likes

@Rene_Krause You mention Serato - Did you import the tracks via your Serato library, or directly into Engine DJ?

It’s best to import directly so that Engine is not influenced by analysis of other software.

Also you say it’s not about variable BPM…but then you say that it’s not dance or electronic music that is affected (which makes variable BPM more likely to be the cause).

Can you give at least one example of a track that has not analysed correctly?

@BeatMaster Danke für die netten Grüße von Hamburg nach Hannover. Wir Grüssen zurück.

1 Like

Thanks for your feedback

the import always runs directly via the hard disk, so no external database in between.

I only mentioned Serato because I had the same problem years ago, just not as severe. I don’t understand why Life titles are more affected, Rock and similar. But as @BeatMaster already wrote, disco genres are strongly affected.

Here is a 4/4 with a continuous beat and yet the life analysis is wrong! And again, neither beat nor key fits. That’s really crap. if you put on life and then drop from 124 to 99 BPM.

We have local hits and the like here that don’t appear in DJ categories in the USA.

For example, Udo Juergens - seventeen years old, blond hair

Engine DJ - BPM 132 Key 1A Serato. - BPM 130 Key 6A How do you deal with the problem if this affects thousands of tracks in your library? Or are there things you can do to minimize or even better solve the problem? Thank you for Help tp all

The key detection app on my phone thinks it’s in Am (8A) and opinion online seems mixed.

Moin again @Rene_Krause,

I will revert on “experience” and I will pick up your example from Udo Juergens and pls let me comment:

BPM Difference between 132 and 130 is extremly small (Don’t ask me today for the differences when we count BPM for 15 seconds with the stop watch and multiplicated by 4). The platter seldom run exactly on 33 1/3 or 45 rpm.

The difference between KEY 1A (Serato) and 6A (Denon) doesn’t surprise me, as there is the “Quintencircle” (have a look at Mixedinkey.com) It’S based on camelot system: If you add 5 to 1 = 6 —> A-Flat-minor und G-minor (As-Moll: G Moll), that matches imo. Or deduct 5 from 6 =1. Best to train a little to detect the differences in sounding. Furthermore the key depends on the speed if you enable the keylock. It will shift, but in the library the original key will be kept constantly.

And an other fact imo is that musicans not using electronical instruments, are always follow the drummer and they are mostly “next” to the beat but never “on” the beat.

This is imo the reason, why old disco tracks (seldom joined by electronical instruments) are “failing” the downbeat. And I think, even the best software can’t analyze 1.000 %o exactly all tracks —a little rest will remain—.

So you have to use your own skills for beatmatching in the cases, when what program ever didn’t hit exactly.

So keep on beatmatching and don’t give up

Brgds BeatMaster

1 Like

Hello :purple_heart:

How do you deal with the problem if this affects thousands of tracks in your library?

I do the analysis manually. I used to think that analysis by software is a cool idea and I used both Serato and Mixed in Key, then I came to the conclusion that no software will give “correct” results in all cases, music is simply often too complex for an analysis that assigns one musical key to an entire track or piece … things may change, things may not be as unambiguous as we could wish for, and finally there is quite some subjectivity involved here. Now I use only Engine DJ and either my piano or a keyboard synth to play along each track to identify the scale - yes, this takes a lot of time, but I consider this a revealing activity.

In case that you do not have keyboard, you can use a virtual keyboard, like this one:

In the case of the Udo Jürgens song, I checked two videos, one seems to be the original version:

After the intro, the verse begins with its harmonic center on C with notes from the C major chord. Playing the C major scale throughout the verse sounds very compatible, so I would say that this is quite clearly C major or 8B on the Camelot Wheel. Here are the chords that I heard:

  • C major: Ein Tag wie jeder …
  • A minor: Ich träum’ von Liiiebe :cupid::two_hearts:
  • F major: … doch eben …
  • G major: … nur ein…
  • E minor: Traum …

And if you take apart each of the chords and the melody of the vocals, all the notes are from the C major scale - which is the same set of notes as the A minor scale, 8A. And maybe that is the reason why PKtheDJ’s app says that it might be 8A … However, if you get a result with an ‘A’, which denotes a minor scale, and the song or track sounds really happy, look at the corresponding key with a ‘B’ on the Camelot Wheel, that may well be a better result.

My guess is, that the software will try to find and identify the notes which are present in the piece to rule out some scales, it may try to identify the harmonic center and decide on one result - but there is quite some subjectivity involved here, as is evident from various discussions online or elsewhere about the key of a given song or track. For us as DJs it may not matter so much, what key the composer actually had in mind, as long as our mix sounds cool.

Let’s look at the song again, because it demonstrates nicely how complicated things can get. After the verse in C major, comes the chorus “17 Jahr’, … blondes Haar…” and I would say it begins with an A major chord, followed by a F# minor chord … then comes another verse “Sie hat mich angelacht …”, which begins with a D♭ major chord melody, … as catchy as the song may sound, it has quite some variation in my opinion. What would I assign here? Probably 8B or C major, because of the first verse and because the intro seems to begin with a C major chord as well. Would I mix this somehow harmonically in a way that takes all the changes into account? :see_no_evil: No, I do not have such mad DJ skillz :sweat_smile:

Often I have a track open in the software and I listen to it and I listen some more and more, and I still cannot find or decide on a key. Often Engine DJ is helpful here, it is like my good friend who offers an opinion, often I agree with it. Some tracks are weird, the harmonic center may not be the expected note from the scale, and modern techno uses more keys than just major and minor keys. Especially acid tracks often have some psychedelic squelchy strangeness in the 303 sequences that corresponds more to a Phrygian scale. Deviating from traditional patterns is one of the aspects that make techno so exciting, new sonic territory and spaces, adventure … here be dragons!

Do I get faster? During the past weeks and months I got quite some practice. So, yes, I get faster, but it still takes some time. I think that as a DJ I should at least listen to each track once, at least to check if the file is ok. Adjusting beat grids takes time too, especially when it is music from a band, a drummer is not a drum machine, small deviations in the tempo will prevent to grid to align perfectly.

The reward of taking the slow route, at least for me, is a deeper understanding of the music and the way tracks are structured, aural training (I do not have perfect pitch), and analysed tracks that mix nicely. I may not analyse thousands of tracks in one go, and it is important to take breaks too, but I am not in a hurry. :relieved: I actually enjoy analysing tracks. I hope this helps you a little on your own journey!

Viele Grüße aus dem Feenwald! :green_heart::herb:

1 Like

Thanks to all the analyzes are all very worth knowing. Now I think I have to put my hand on many places here, although I thought that with the number of songs it will be difficult. Finding the key is not easy, I understand that, but why is the beat grid not right whether the songs have clear pulses?

Thanks a lot for your help

The example you gave is from the 1960s so it would be expected for the tempo to vary, because the drums are played by a human.

You also mentioned that dance and electronic music is not affected, which would be expected.

Typically, BPM analysis in most software gives an average of the tempo, and applies a fixed (equally spaced) grid.

Unfortunately this will not match exactly with music which is either played live, or from an era where small speed variations on equipment can creep in - i.e. tape machines.

The only software (I’m aware of) which is smart enough to analyse variable tempos accurately is the recently released Algoriddim DJay Pro 5 which uses a system called Fluid Beatgrid™.

Hopefully other companies will follow suit, because it’s a big time saver.

3 Likes

The pulse may be clear to you, for your brain it is easy to differentiate between a kick drum, a bass, a snare drum, a trumpet, a singer … the software will see a waveform and has to find a rhythmic pattern, but the pattern may not be that regular or it may find a pattern in a wrong place. In a simple case, the grid may be aligned with the off-beat, and we can shift it. Sometimes the software simply fails to find the pattern … and sometimes it finds a plausible pattern, where I did not :blush:

By doing the analysis, I have found tracks which speed up a little over time, tracks which drift and swerve, and what I call “Frankenstein” tracks, which are pieced together from several parts, each with a slightly different tempo … yes, it is annoying and placing beat marks manually is what I like least, but we may eventually see better tools for beat grid analysis in the future. :four_leaf_clover:

1 Like

Denon really needs to set a few of their software engineers to the task of improving beat grid detection. Engine is embarrassingly bad at getting beat grids right and it’s one of the first things people notice when they use the software. I mix four-to-the-floor EDM music which should be the absolute easiest thing to get right and Engine still screws it up like 1/3 of the time. I’d go as far as to say that you could fix half the issues with beatgrid analysis by just defaulting to using whole integer BPMs – for some reason stuff like 149.97 or 128.03 BPM comes up wayyyy too often (and yes it’s wrong and when I set to integer BPMs it fixes it). Nobody makes music in those BPMs, guys. Or if they do it’s like .00000001% of songs out there. And this isn’t some unsolvable problem – Traktor is miles better than Engine in this regard. For everything Denon does right, it boggles my mind that you guys haven’t sat down and put in one single line of code to round BPMs and fix half your analysis issues. Sorry if I’m being harsh but like – I’d like to see the analysis stop being bad soon, lol.

1 Like

Maybe not deliberately, no - but just because the tempo in the artist’s DAW says 128, it doesn’t necessarily mean that it is exactly 128.

Then between leaving the DAW and getting to the DJ, it could have been shifted by maybe being re-encoded into various formats for retail, being mastered by the label, being run through e.g. Platinum Notes by DJs…

If it’s older music, all sorts of other things could have affected the tempo, like analogue tech, tape machines, off centre vinyl pressings…

2 Likes

Maybe not deliberately, no - but just because the tempo in the artist’s DAW says 128, it doesn’t necessarily mean that it is exactly 128.

Then between leaving the DAW and getting to the DJ, it could have been shifted by maybe being re-encoded into various formats for retail, being mastered by the label, being run through e.g. Platinum Notes by DJs…

I’ve literally never seen a song made in a DAW be even a fraction of a BPM off of whole. It’s not a thing that ever happens, or if it does, it’s so utterly rare that I’d happily deal with that the one or two out of thousands of times it occurs, so that I don’t have to deal with fixing hundreds of tracks that are integer BPM but Engine thinks aren’t.

If it’s older music, all sorts of other things could have affected the tempo, like analogue tech, tape machines, off centre vinyl pressings…

Well you’re never going to get a good BPM read on those genres anyway, so it’s silly trying to hit a target that you inevitably miss, causing you to screw up on the vast majority of music out there. Like nobody is going to fault Engine for failing to analyze a track that was pressed onto vinyl and then tape recorded. But Engine should absolutely be faulted for constantly missing on modern digital music with silly mistakes like thinking songs are 149.97 bpm.

Occasionally I get non-integer BPM results too, but often they are correct. Tracks made with analogue gear are not always exact up to integer accuracy, especially older techno tracks often drift. Sequencers are not always so accurate, a MIDI clock can be wonky, and if the track is based on using a looper quite a lot of things may be possible with respect to the tempo … analogue gear is sensitive to temperature, maybe it was getting hot in the studio. :sunglasses:

I would say that if the resolution of the analysis were high enough, we would see that almost no track has an integer BPM. A producer may start with a kick drum, a BPM value, and then add all the other things. Or they may start with a sample, something that loops nicely and sounds groovy at a given tempo, but the BPM may not be an integer.

Occasionally I get non-integer BPM results too, but often they are correct. Tracks made with analogue gear are not always exact up to integer accuracy, especially older techno tracks often drift. Sequencers are not always so accurate, a MIDI clock can be wonky, and if the track is based on using a looper quite a lot of things may be possible with respect to the tempo … analogue gear is sensitive to temperature, maybe it was getting hot in the studio. :sunglasses:

I would say that if the resolution of the analysis were high enough, we would see that almost no track has an integer BPM. A producer may start with a kick drum, a BPM value, and then add all the other things. Or they may start with a sample, something that loops nicely and sounds groovy at a given tempo, but the BPM may not be an integer.

Alright guys, I’m sorry but these are some super silly takes. For every one song that has 149.97 BPM, ten thousand have 150 BPM. The analysis shouldn’t get hundreds of 150 BPM songs wrong by .03 BPM in an attempt to once in a blue moon nail the 149.97 song.

Can you guys name me some of these songs where the legit version is .03 BPM off whole? Because I am super skeptical.

Inmusic should buy Algoriddim and put DjPro analyse technology in Prime series. That would be fabulous ! And be the end of the “other brand”

While I agree that BPM analysis is not flawless, I want to give some silly examples, where the BPM is off by a small amount, metronome checked:

  • Cpt. Nik - Charabia (164.98 BPM)
  • Benny-Bien4cid - Ritaline (160.04 BPM)
  • Doc Tom - Goon Song (184.97 BPM)
  • Marcus Henriksson - Fire Of Water (Don Pepe & David Laake Remix) (127.02 BPM)
  • Beat Bizarre - Get Down (132.96 BPM)

When I find tracks like these, I try out an integer BPM value as well, but in these cases it was just not right and the beat and the grid were drifting apart … as long as the beat is regular, I do not see much of a problem here. Tracks with a drifting tempo are a problem, because it is tedious to go through the track and set the beat marks manually for the grid to align with the beat - you will get a grid with a varying BPM value, an example:

  • Jammin’ Unit - Temptation (Force Inc.) 156.48, 155.85, 155.96 and so on, drifts like crazy

Some people find this a little unsettling …

hm… i will check that

I’m totally blown away. It pulls my the shoes out. :slight_smile: Now I know why it bothered me so much. @Guetto_David is absolutely right, Algoriddim DjPro analysis technology is what I always had in the back of my mind. And now just started the app and ran my problem kids Disco 12’ and Ah hA, there they are my expectations. That’s exactly how it should look.

I took two screenshots and you can see that the algorithm even adjusts the BPM on the fly. :open_mouth: see picture 1 and picture 2 right player! BPM display the same song.

1

2

As you can see, this is the only way Stem analysis can work properly.
But the dear developers make massive use of the graphics card to do the calculations. So it gets loud. (fan)

comparison:

DJ Pro slave - Just a Touch of Love (12" Lp Version) 109,2 BPM Key 7A but the track starts with 111 BPM the big difference is that the beat grid is always exactly on the hi-hat.

engine DJ slave - Just a Touch of Love (12" Lp Version) 109 8A

it’s amazing

Best René

2 Likes