Helpful ReplyRecent experience when extracting tempo

Author
Rob[at]Sound-Rehab
Max Output Level: -47 dBFS
  • Total Posts : 2819
  • Joined: 2011/02/03 04:31:35
  • Location: Sound-Rehab, Austria
  • Status: offline
2017/03/12 09:10:25 (permalink)

Recent experience when extracting tempo

One of my recent tasks was to link a band recording (8 songs, 20-22 tracks per song) to a tempo grid so that the production could use tighly snapped synths of strong rhythmic content.  I really had a mixed experience with Sonar and want to share that here, partly hoping for tips that get me further, partly hoping that this picks up momentan as I believe this to be one of the areas where Sonar still lacks considerably.
 
Here's the story:  Songs were tracked against a click track / pre-production synth playback (from 44.1 kHz bands live system), all signals were recorded at 96 kHz with the DAW (Sonar) only acting as "tape machine" during the recording session, in which the band recorded 3-4 takes of each song before moving to the next. The idea was to capture as much of the live band character as possible and later select the best take (and use other takes for comping if there is something that needs to be fixed). So far so good. Most of the songs were indeed derived from the single best take. Edits were straight forward, comping worked fine with takes organised as clip groups in Sonar.  The fun started when realizing that 131 BPM from the click track machine is not exactly 131 BPM on the recording machine due to different clocks (more like 130.99). So over the course of a 5 min song a slight drift would develop, meaning tempo synced VSTi synths that are tighly snapped to the audio at the beginning go out of sync at the end. I did not want to "quantize" the band recording and suck the live out of it, so I decided to generate tempo grids for each song that would slighly adjust the tempo and maintain the sync.  Along the lines I intended to snap the synths to subtle variations you get from a real drummer.

Yet, here the frustrating experience started. Virtually every song required a different approach. First I was surprised how well audio snap worked (detect transients => QC visually & adjust => apply clip tempo to project). Yet in the next song it would not work at all. Whatever it extracted did not sync properly, so something in the calculation was simply off.  The result it produced did not reflect the transients which were the input to the calculation.  Using "Edit Clip Map" makes you manually link the first beat of every bar (i.e. 100 mouse clicks for 100 bars) to have it work in one song, not in the next.  Then you try to manually snap every mapped transient (every quarter note), which would mean 400 mouse clicks, but after the first 30 you test and realize that it cannot do this simple grammar school exercise of computing the time between transients and translate to a BPM change (because it's either off or comes back with the message that it cannot compute it??? how come???).

I also tried Melodyne tempo extraction (which works fine in Melodyne i.e. it nicely detects the tempo), but has some serious glitches in Sonar e.g. if you do not split exactly at the first transient and snap the clip to the exact start of a measure, it won't work.  If you do mentioned prep steps you get a nice tempo sync, but it creates so many tempo changes in one bar, that VSTis produce really poor sound (you get ditigal distortion because VSTis cannot keep up with the overload of tiny tempo changes). There are no options to control how the extracted tempo is to be applied (e.g. map to every quarter, map to every bar, thus re-scale the Melodyne tempo grid for practical purposes)
 
So to get my task done (get a full song tempo sync but proper synth sound quality), I needed a tempo grid with ideally only 1 tempo change per bar, the change to be at the beginning of the bar that gets me exactly to the 1st beat in the next bar.  Simple but hardly impossible to achieve.  Likewise if you want to link to the back beat (2s+4s) because that's what drives the song and the 1 is no always there ...
 
OK, this post is running long ... I made some strange observations in this entire procress. One also being the more subtle the changes need to be the less likely Sonar will be able to get them (I actually ended up painting the tempo grid with the mouse which turned out to be best and fastest option in the last 2 songs !!!), which makes me conclude that both options (audio snap, Melodyne) are only half baked at best, meaning that they do work somehow and between them you can get most of what you need to get done, but it usually comes with a LOT of frustration and a massive time investment, which are two things you don't need when you need to make deadlines.
 
 

GOOD TUNES LAST FOREVER
  +++   Visit the Rehab   +++
 
DAW: Platinum/X3e, win10 64 bit, i7-3930K (6x3.2GHz), Asus Sabertooth X79, 32 GB DDR3 1600MHz, ATI HD 5450, 120 GB SSD OCZ Agility3, 2x 1TB WD HDD SATA 600
Audio-Interface: 2x MOTU 1248 AVB, Focusrite OctoPre, (Roland Octa-Capture)   Control-Surface: VS-700C 
VSTi: WAVES, NI K10u, FabFilter, IK, ... (too many really) 
#1
chuckebaby
Max Output Level: 0 dBFS
  • Total Posts : 13146
  • Joined: 2011/01/04 14:55:28
  • Status: offline
Re: Recent experience when extracting tempo 2017/03/12 15:03:50 (permalink)
I do a lot of tempo synching myself. and for the most part, your correct.
Melodyne and Audiosnap work great on something but not all things.
as much as I love Melodyne's tempo mapping feature, it takes set up time and choosing the correct algorithm.
you also need to sometimes either leave a space or have the content start exactly at the beginning of the project which is a no no in my book. I always start projects at 2:01.
 
I almost always synch the grid by drawing a tempo map. yes it takes time but I have tried this in other DAW's and have achieved less than stellar results. So its not just Sonar on this one. Its the day and age we live in. Some tools need to be taken for what they are... Too far a head of their time .
 
If we were to go back 10 years ago and someone said, I have a tool that will allow you to do some tempo mapping but it takes some time. We would be doing back flips asking for it. Now we take a lot for granite and expect better results. So some of it is a sign of the times we live in.
 
I would line up the grid by mapping it, if you used a metronome to record these songs it should be a piece of cake (even with tempo changes included). I do a lot of analog tempo matching. That in itself is a huge PIA because the old capstan motors on these analog tape machines have so much drift (some times as much as  0.3% - 1.1%) That not only are you mapping the song, your having to deal with very subtle changes within the song. 
 

Windows 8.1 X64 Sonar Platinum x64
Custom built: Asrock z97 1150 - Intel I7 4790k - 16GB corsair DDR3 1600 - PNY SSD 220GB
Focusrite Saffire 18I8 - Mackie Control
   
#2
bitman
Max Output Level: -34 dBFS
  • Total Posts : 4105
  • Joined: 2003/11/06 14:11:54
  • Location: Keystone Colorado
  • Status: offline
Re: Recent experience when extracting tempo 2017/03/12 15:24:19 (permalink)
It would be nice if the tempo map could be thinned like controller data.
 
Averaging would be nice too. - I fixed this problem for myself by recording the reference track intended for tempo extraction to a drum machine pattern to keep me more "around" the desired bpm for 3 minutes and 30 seconds. Because without it I was surprised at how much of a musical spaz I was racing and dragging with just a guitar and a record light. - psych psych psych!
#3
brundlefly
Max Output Level: 0 dBFS
  • Total Posts : 14250
  • Joined: 2007/09/14 14:57:59
  • Location: Manitou Spgs, Colorado
  • Status: offline
Re: Recent experience when extracting tempo 2017/03/12 16:17:22 (permalink) ☄ Helpfulby Rob[at]Sound-Rehab 2017/03/12 22:50:57

Rob[atSound-Rehab]
So to get my task done (get a full song tempo sync but proper synth sound quality), I needed a tempo grid with ideally only 1 tempo change per bar, the change to be at the beginning of the bar that gets me exactly to the 1st beat in the next bar.



Yes, that's usually what I aim for when syncing a live performance that's basically a fixed tempo, but not recorded to a click (or not exactly matching due to clock difference as in your case). Often you can go 4-8 bars without a change, depending on how tight you need it to be. For this and other reasons, I still prefer doing it manually, using 'Set Measure/Beat at Now' (Shift+M) to tell SONAR where the downbeats fall. In the case that the song starts at 1:01:000, and already has the average tempo nearly matched, the process is pretty straight-forward.
 
Start by setting the Now time (snap disabled) to the last downbeat in the piece, and 'setting' the measure/beat for that absolute Now time. SONAR will re-calculate the initial tempo to make that beat hit that time as nearly as possible within the limits of .01 BPM tempo resolution. If the song has a partial 'pickup' measure, you'll want to start by dragging the audio to align the first downbeat to 2:01:000, and setting that point.
 
Then start listening from the beginning with the playback metronome enabled, and snap additional downbeats as necessary to get and keep things tight throughout. Using keyboard shortcuts for start/stop playback, prev/next measure, tab to transient and horizontal zoom,  you should be able to work through the piece quite quickly. For me , this generally goes much more quickly than trying to massage AS transient markers to the point that Follow Project will work correctly, and even then, Follow Project does not place tempo changes exactly on bar lines (last worked correctly in 8.5 IIRC). Set Measure/Beat will place the tempo change exactly on the beat you specify, and allows you to 'set' any fraction of a beat anywhere in a measure when a measure has a rit. or a pause or when the audio that you're snapping to just doesn't have a transient on the beat that you want to snap.

SONAR Platinum x64, 2x MOTU 2408/PCIe-424  (24-bit, 48kHz)
Win10, I7-6700K @ 4.0GHz, 24GB DDR4, 2TB HDD, 32GB SSD Cache, GeForce GTX 750Ti, 2x 24" 16:10 IPS Monitors
#4
jpetersen
Max Output Level: -61 dBFS
  • Total Posts : 1499
  • Joined: 2015/07/11 20:22:53
  • Status: offline
Re: Recent experience when extracting tempo 2017/03/12 16:46:46 (permalink)
Nearly all my projects are like this.
 
If the band was tight enough and stuck to their click, then all I do is slip-stretch all tracks together to the correct length.
 
The hotkeys changed recently. Now you press Ctrl + Shift, then move the right side of the clip to the desired length.
 
The bass drum is a good guide. Get it to line up with the bar lines in Sonar. Shift and/or stretch the clips in unison to get it right across the full length of the song.
#5
brundlefly
Max Output Level: 0 dBFS
  • Total Posts : 14250
  • Joined: 2007/09/14 14:57:59
  • Location: Manitou Spgs, Colorado
  • Status: offline
Re: Recent experience when extracting tempo 2017/03/12 16:57:54 (permalink)
Setting the project to the audio is preferable to stretching if you're okay with the loose timing. Even minimal audio stretches will leave audible artifacts.

SONAR Platinum x64, 2x MOTU 2408/PCIe-424  (24-bit, 48kHz)
Win10, I7-6700K @ 4.0GHz, 24GB DDR4, 2TB HDD, 32GB SSD Cache, GeForce GTX 750Ti, 2x 24" 16:10 IPS Monitors
#6
Rob[at]Sound-Rehab
Max Output Level: -47 dBFS
  • Total Posts : 2819
  • Joined: 2011/02/03 04:31:35
  • Location: Sound-Rehab, Austria
  • Status: offline
Re: Recent experience when extracting tempo 2017/03/12 22:54:38 (permalink)
chuckebaby
I almost always synch the grid by drawing a tempo map. yes it takes time but I have tried this in other DAW's and have achieved less than stellar results. So its not just Sonar on this one. Its the day and age we live in. Some tools need to be taken for what they are... Too far a head of their time .




So I'm not alone in this strange world of hand drawn tempo maps. 
It just felt really funny that after 6 different hi-tech software approaches that took ages to get working, I had to realize that painting it by hand (zoomed into the kick track and snap set to one bar) is about 100x times quicker
 
brundlefly
 
Start by setting the Now time (snap disabled) to the last downbeat in the piece, and 'setting' the measure/beat for that absolute Now time. SONAR will re-calculate the initial tempo to make that beat hit that time as nearly as possible within the limits of .01 BPM tempo resolution. If the song has a partial 'pickup' measure, you'll want to start by dragging the audio to align the first downbeat to 2:01:000, and setting that point.
 
Then start listening from the beginning with the playback metronome enabled, and snap additional downbeats as necessary to get and keep things tight throughout. Using keyboard shortcuts for start/stop playback, prev/next measure, tab to transient and horizontal zoom,  you should be able to work through the piece quite quickly. For me , this generally goes much more quickly than trying to massage AS transient markers to the point that Follow Project will work correctly, and even then, Follow Project does not place tempo changes exactly on bar lines (last worked correctly in 8.5 IIRC). Set Measure/Beat will place the tempo change exactly on the beat you specify, and allows you to 'set' any fraction of a beat anywhere in a measure when a measure has a rit. or a pause or when the audio that you're snapping to just doesn't have a transient on the beat that you want to snap.




This is GREAT advice. Tabing thru transients using SHIFT+M to adjust the tempo. This is so much quicker than anything you could do with audio snap ... and you can snap to the offbeat as well.
 

GOOD TUNES LAST FOREVER
  +++   Visit the Rehab   +++
 
DAW: Platinum/X3e, win10 64 bit, i7-3930K (6x3.2GHz), Asus Sabertooth X79, 32 GB DDR3 1600MHz, ATI HD 5450, 120 GB SSD OCZ Agility3, 2x 1TB WD HDD SATA 600
Audio-Interface: 2x MOTU 1248 AVB, Focusrite OctoPre, (Roland Octa-Capture)   Control-Surface: VS-700C 
VSTi: WAVES, NI K10u, FabFilter, IK, ... (too many really) 
#7
VariousArtist
Max Output Level: -63 dBFS
  • Total Posts : 1397
  • Joined: 2003/11/07 15:03:09
  • Location: London, UK & California, USA
  • Status: offline
Re: Recent experience when extracting tempo 2017/03/15 04:10:43 (permalink)
This all sounds very familiar to me. It's not an easy task by any stretch of the imagination (or the clip).

Subtleties will trip you up. So will one misplaced beat marker. And if you decide to cut and paste a section from one song take into another, then it can better to do all this without any tempo maps and only attempt to do the tempo grid *after* you have bounced the clips.

I save frequently with different version numbers in the file name and include descriptions like "v1.4 just before bounce", etc. And I generally try to make all the clips on each track exactly the same length starting from the beginning. That way I avoid any surprises when I start working on tempo maos and so on.

Sonar can be really helpful and powerful, but you can easily get into a world of trouble. The above represent my lessons learned but I'd like to see this whole area improved further.
#8
Rob[at]Sound-Rehab
Max Output Level: -47 dBFS
  • Total Posts : 2819
  • Joined: 2011/02/03 04:31:35
  • Location: Sound-Rehab, Austria
  • Status: offline
Re: Recent experience when extracting tempo 2017/03/15 05:45:51 (permalink)
VariousArtist
Sonar can be really helpful and powerful, but you can easily get into a world of trouble. The above represent my lessons learned but I'd like to see this whole area improved further.


I very much agree. Some of the improvements would be rather simple, either e.g.
- tempo map import from / export to flat ascii or csv files
- copy paste tempo to spreadsheet programs
- quantizing the tempo grid i.e. recompute continuous change extracted by Melodyne ARA functionality to one tempo change per quarter or one per beat

GOOD TUNES LAST FOREVER
  +++   Visit the Rehab   +++
 
DAW: Platinum/X3e, win10 64 bit, i7-3930K (6x3.2GHz), Asus Sabertooth X79, 32 GB DDR3 1600MHz, ATI HD 5450, 120 GB SSD OCZ Agility3, 2x 1TB WD HDD SATA 600
Audio-Interface: 2x MOTU 1248 AVB, Focusrite OctoPre, (Roland Octa-Capture)   Control-Surface: VS-700C 
VSTi: WAVES, NI K10u, FabFilter, IK, ... (too many really) 
#9
chuckebaby
Max Output Level: 0 dBFS
  • Total Posts : 13146
  • Joined: 2011/01/04 14:55:28
  • Status: offline
Re: Recent experience when extracting tempo 2017/03/15 12:27:35 (permalink)
I really would love to see a tempo track.
Something you could automate using nodes the same way you use automation.
The tempo view in Sonar to me is too dated. it does the job but I think it needs an upgrade.
So a tempo track and a markers track are 2 things on my wish list. these would allow tempo and marker data to captured with data vs. needing to have the project, which is another plus.

Windows 8.1 X64 Sonar Platinum x64
Custom built: Asrock z97 1150 - Intel I7 4790k - 16GB corsair DDR3 1600 - PNY SSD 220GB
Focusrite Saffire 18I8 - Mackie Control
   
#10
mettelus
Max Output Level: -22 dBFS
  • Total Posts : 5321
  • Joined: 2005/08/05 03:19:25
  • Location: Maryland, USA
  • Status: offline
Re: Recent experience when extracting tempo 2017/03/15 20:55:22 (permalink)
+1 for Shift-M (Set measure/beat at now). When all tracks are in sync this is a very quick workflow.

ASUS ROG Maximus X Hero (Wi-Fi AC), i7-8700k, 16GB RAM, GTX-1070Ti, Win 10 Pro, Saffire PRO 24 DSP, A-300 PRO, plus numerous gadgets and gizmos that make or manipulate sound in some way.
#11
Jump to:
© 2024 APG vNext Commercial Version 5.1