Audiosnap transient detection is much improved.

Author
Dave Modisette
Max Output Level: 0 dBFS
  • Total Posts : 11050
  • Joined: 2003/11/13 22:12:55
  • Location: Brandon, Florida
  • Status: offline
2011/03/23 20:47:48 (permalink)

Audiosnap transient detection is much improved.

When I opened the Audiosnap pallette I was a bit disappointed.  I believe it was Ryan that hinted that Audiosnap was getting a major rework and everything looked the same.  I really expected some sort of gating and filtering added that would make drum replacement easier.  The old Audiosnap was all over the place when detecting transients which cause as much hand editing as Audiosnap produced.  Transient markers would be all over the place even on the most defined transients.

Well, it looks like the transient detection has definitely been improved even if the additional tools I hoped for weren't integrated into the interface.

That said, there were two improperly detected in a thirty something bar long Superior Drummer frozen track.  I know, "what's two bad ones?"  Yeah, I know but I expect a pop-y snare drum to never be mis-detectec.  You can't get more of a transient than this.

So I mean this as an encouragement to keep banging on Audiosnap till it is perfect.

Here is the missed transient to show how easy it should have been to detect.  (All the others were perfect - even on drum tracks that had more subtle transients.)





Dave Modisette ... rocks a Purrrfect Audio Studio Pro rig.

http://www.gatortraks.com 
My music.
... And of course, the Facebook page. 
#1

21 Replies Related Threads

    Noel Borthwick [Cakewalk]
    Cakewalk Staff
    • Total Posts : 6475
    • Joined: 2003/11/03 17:22:50
    • Location: Boston, MA, USA
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/23 21:37:15 (permalink)
    Can you send us the wave file for that segment where the transient is not detected? It would be useful to analyze to see why that one transient was missed.

    Noel Borthwick
    Senior Manager Audio Core, BandLab
    My Blog, Twitter, BandLab Profile
    #2
    Dave Modisette
    Max Output Level: 0 dBFS
    • Total Posts : 11050
    • Joined: 2003/11/13 22:12:55
    • Location: Brandon, Florida
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/23 21:47:54 (permalink)
    Noel Borthwick [Cakewalk
    ]

    Can you send us the wave file for that segment where the transient is not detected? It would be useful to analyze to see why that one transient was missed.


    Unfortunately, I unfroze the track.  I'll try to see if I can get the same result.

    Dave Modisette ... rocks a Purrrfect Audio Studio Pro rig.

    http://www.gatortraks.com 
    My music.
    ... And of course, the Facebook page. 
    #3
    A1MixMan
    Max Output Level: -58 dBFS
    • Total Posts : 1706
    • Joined: 2003/11/19 16:15:11
    • Location: SunriseStudios
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/23 22:04:29 (permalink)
    Noel Borthwick [Cakewalk
    ]

    Can you send us the wave file for that segment where the transient is not detected? It would be useful to analyze to see why that one transient was missed.

    How's that for service?
     
    Whatever problems there are, no one can say that these guys aren't trying...
    post edited by A1MixMan - 2011/03/23 22:06:18

    A1
    #4
    StarTekh
    Max Output Level: -55 dBFS
    • Total Posts : 2007
    • Joined: 2004/03/09 12:02:20
    • Location: Montreal
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/23 22:12:34 (permalink)

     > Up and Above and Beyond..rings a bell..jon <
    #5
    revsnd
    Max Output Level: -86 dBFS
    • Total Posts : 206
    • Joined: 2004/01/12 10:50:31
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/23 22:16:25 (permalink)
    I noticed it generally much better tonight also.  Keep it going though:)
    #6
    Dave Modisette
    Max Output Level: 0 dBFS
    • Total Posts : 11050
    • Joined: 2003/11/13 22:12:55
    • Location: Brandon, Florida
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/23 22:35:17 (permalink)
    Noel Borthwick [Cakewalk
    ]

    Can you send us the wave file for that segment where the transient is not detected? It would be useful to analyze to see why that one transient was missed.


    I was able to reproduce the same result and I have submitted a problem report with the file attached.

    I've also had a crash in the process and I have also filed a problem report for that too and included a mini dump.

    At this time X1b is hung and not responding and I expect that when I close it I will likely get another BSOD.  If you need that additional minidump I can upload that at another time.

    I suspect the crash is due to to a Synth Edit free softsynth (Organized Trio) in this default project template.  SynthEdit softsynths have always been dodgy but I happen to like this one and most times it behaves itself.

    EDIT: No crash this time.  There was a window open waiting on a response and it was covered by another open window.
    post edited by Mod Bod - 2011/03/23 22:37:28

    Dave Modisette ... rocks a Purrrfect Audio Studio Pro rig.

    http://www.gatortraks.com 
    My music.
    ... And of course, the Facebook page. 
    #7
    Noel Borthwick [Cakewalk]
    Cakewalk Staff
    • Total Posts : 6475
    • Joined: 2003/11/03 17:22:50
    • Location: Boston, MA, USA
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/23 22:50:13 (permalink)
    Thanks

    Noel Borthwick
    Senior Manager Audio Core, BandLab
    My Blog, Twitter, BandLab Profile
    #8
    panup
    Max Output Level: -50 dBFS
    • Total Posts : 2538
    • Joined: 2006/05/23 09:34:35
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/23 23:09:44 (permalink)
    I have similar results with X1b AudioSnap.
    Live Kick drum track miced inside with AKG D112 in studio, only drums in the room (no bass spill etc). Looks like this:



    EDIT:
    Project that contains Kick track with default AudioSnap setting is here: x1b audiosnap01 project.rar
    post edited by panup - 2011/03/23 23:13:54
    #9
    brundlefly
    Max Output Level: 0 dBFS
    • Total Posts : 14250
    • Joined: 2007/09/14 14:57:59
    • Location: Manitou Spgs, Colorado
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/23 23:48:21 (permalink)

    The old Audiosnap was all over the place when detecting transients which cause as much hand editing as Audiosnap produced.  Transient markers would be all over the place even on the most defined transients. Well, it looks like the transient detection has definitely been improved even if the additional tools I hoped for weren't integrated into the interface.



    I'd be really surprised if they did that without mentioning it. I'll check it out, but I suspect it's placebo effect.


    As for the one misplaced transient marker you showed, I'll wager if you zoom the amplitude on that transient, you'll find a discontinuity or bleed from another track that triggered the early placement.


    You mentioned gating and filtering for drum replacement. It's not built in to AS, or course, but it is easy and helpful to apply gating and/or EQ to a clone of a drum track before enabling AS to get more precise transient detection. It's especially helpful when you get a lot of these early transient detections which I have seen before with some drum tracks. 


    EDIT: If you've already done it, I'd be interested to see screenshots of the same track with AS applied in X1 vs. 8.5.3 that show a difference in transient detection.
    post edited by brundlefly - 2011/03/23 23:50:50
    #10
    jamescollins
    Max Output Level: -76 dBFS
    • Total Posts : 747
    • Joined: 2009/04/06 19:33:06
    • Location: Perth, Australia
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/24 03:28:18 (permalink)
    That's great that detection seems to be improving - has anyone noticed this for live drums, or just samples? I can't test right now as I'm away, but any improvement in transient detection of live audio would be VERY welcome!

    I'll have three fingers of Glenlivet, with a little bit of pepper... and some cheese.
     
    allthekingsmen.band
    jamescollinsmusic.com
    #11
    Dave Modisette
    Max Output Level: 0 dBFS
    • Total Posts : 11050
    • Joined: 2003/11/13 22:12:55
    • Location: Brandon, Florida
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/24 07:37:37 (permalink)
    As for the one misplaced transient marker you showed, I'll wager if you zoom the amplitude on that transient, you'll find a discontinuity or bleed from another track that triggered the early placement.

     
    There was a minimum of two in a thirty bar section.  I scaled the waveform  prior to my post to see if there was some little niggle that was triggering the detection.  There wasn't one.
     
    Regardless of that;
     
    1) The Threshold and Resolution controls should be able to filter out any background bleed in a cut and dry, multitrack snare drum track.  The transients are too defined to not be able to accomplish that.
     
    2) The ratio of wrong guess to correct guesses is irrevelant if this is supposed to be a professional app (especially in this particular situation.)  It's all about confidence in your tools here.  If you create a series midi events for the purposes of triggering a sequencer, you shouldn't have to waste your time resources checking to see if the tool actually performed as expected.  Your client isn't paying you to do that and Cakewalk certainly isn't kicking you back any rebates for it either.
     
    3) Creation of midi events is included in Audiosnap.  This is naturally going to suggest drum replacement.  I've been harping (and suggesting via feature requests) that just a few more controls and some user preset saving features could put them in the game.  When AudioSnap detects the transients correctly, it is really easy to create layered (or replaced) drum tracks.

    Dave Modisette ... rocks a Purrrfect Audio Studio Pro rig.

    http://www.gatortraks.com 
    My music.
    ... And of course, the Facebook page. 
    #12
    formermasterbaker
    Max Output Level: -90 dBFS
    • Total Posts : 4
    • Joined: 2009/07/25 07:06:43
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/24 07:53:30 (permalink)
    I'm having a lot more love with AudioSnap these days.  I limit my use strictly to percussion tracks because these are the only ones that can render in real-time.  If we're ever able to freeze individual clips, I'll use it everywhere.  But that's just details :)  For now, doing percussion edits gets me the 90/10 functionality and the tools that create the stretches are working beautifully for me now.   I've convinced a few friends to give A/S a second look after they gave up in previous versions.  I may have to do a private clinic tour.  Too bad they don't live anywhere tropical.
    #13
    FastBikerBoy
    Forum Host
    • Total Posts : 11326
    • Joined: 2008/01/25 16:15:36
    • Location: Watton, Norfolk, UK
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/24 08:24:17 (permalink)
    I'm no audiosnap expert having only used it a couple of times before but I've just been using it on a guitar line and it worked great. Definitely improved and easier to use than my last effort in 8.3, bbut I don;t think I was pushing the envelope by any stretch of the imagination.

    Now if I could only get Sonar to record the guitar parts in time in the first place I wouldn't need it. 
    #14
    brundlefly
    Max Output Level: 0 dBFS
    • Total Posts : 14250
    • Joined: 2007/09/14 14:57:59
    • Location: Manitou Spgs, Colorado
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/24 14:17:47 (permalink)
    There was a minimum of two in a thirty bar section.  I scaled the waveform  prior to my post to see if there was some little niggle that was triggering the detection.  There wasn't one.   Regardless of that...


    I'm not saying there's no room for improvement. Just mentioning a common cause for transient markers being located too early. The challenge for the algorithm is to figure out where the transient actually starts, since even drum transients can have an attack that spans a couple milliseconds.

    And since placing a marker too late could play hell with the attack sound if part of it gets stretched, I think the algorithm is designed to err on the early side. This doesn't explain your example. which looks pretty extreme, but it could account for with Panup's example where the marker appears to be only 8-10 ticks early (4-5ms at 125 BPM).







    #15
    brundlefly
    Max Output Level: 0 dBFS
    • Total Posts : 14250
    • Joined: 2007/09/14 14:57:59
    • Location: Manitou Spgs, Colorado
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/24 14:27:02 (permalink)
    FBBDefinitely improved and easier to use than my last effort in 8.3



    There were huge changes between 8.3 and 8.5, and more from 8.5.0 to 8.5.3, but not so much (if any) from 8.5.3 to X1.


    I did a little testing on  few clips, and did not see even a one-sample difference in where transients were located between 8.5.3 and X1b.


    I did notice in the process, however, that the problem with spurious "User" markers (solid diamond handles) being added on beats that don't have corresponding audio transients persists in X1b.   Easily fixed by selecting all User markers and deleting them right after  enabling AS, but it's one of those little PITA bugs that would be nice to be rid of.


     
    #16
    FastBikerBoy
    Forum Host
    • Total Posts : 11326
    • Joined: 2008/01/25 16:15:36
    • Location: Watton, Norfolk, UK
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/24 14:39:59 (permalink)
    There were huge changes between 8.3 and 8.5, and more from 8.5.0 to 8.5.3, but not so much (if any) from 8.5.3 to X1.

    Never used 8.5 for long, I took advantage of the BOGOF offer so effectively went from 8.3 -> X1
    #17
    brundlefly
    Max Output Level: 0 dBFS
    • Total Posts : 14250
    • Joined: 2007/09/14 14:57:59
    • Location: Manitou Spgs, Colorado
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/03/24 15:01:59 (permalink)
    Noel Borthwick [Cakewalk
    ]

    Can you send us the wave file for that segment where the transient is not detected? It would be useful to analyze to see why that one transient was missed.

    I don't suppose Noel is still tuned into this thread, but thought I'd mention that the Kick and Snare tracks from the old Guilty demo by Unified Tribe are good examples of transient misdetection - due, at least in part, to long "pre-attacks" on drum transients that look to have been caused by a hardware gate/compressor with a long attack being applied to the analog signal before it every got recorded.






    #18
    sdpate67
    Max Output Level: -84 dBFS
    • Total Posts : 344
    • Joined: 2008/03/09 09:59:21
    • Location: Charlottetown, PEI
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/04/03 23:13:57 (permalink)
    Does anyone use the auto quantize or are you moving the transient markers by hand?

    I get weird results with quantize but it works ok lining up the transients on a bass line, except you have to demote 1/8th notes on turn arounds.

    Asus i7-760 Win 8.1/ Sonar Platinum / Lynx Aurora 16 AES16 / Mackie MCU Pro XT C4 / Millennia Media STT1 x 2 TD-1/ UAD-2 Quad x 2 / Neumann O-300 O-810 U87 KM184 x 2 / Shure 57/58
    Reverbnation

    NJN Network
    #19
    Dave Modisette
    Max Output Level: 0 dBFS
    • Total Posts : 11050
    • Joined: 2003/11/13 22:12:55
    • Location: Brandon, Florida
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/04/04 07:46:17 (permalink)
    sdpate


    Does anyone use the auto quantize or are you moving the transient markers by hand?

    I get weird results with quantize but it works ok lining up the transients on a bass line, except you have to demote 1/8th notes on turn arounds.
     
    I've been fond of saying that anything that "automatically" does something usually does it automatically wrong.  I usually move the transients by hand as I line problems up with another instrument that it relates to.  For example, a bass guitar with a kick.
     
    I may at some times use an automatic mode to align a choppy rhythm guitar part and quantize with a percentage to a grid.


    Dave Modisette ... rocks a Purrrfect Audio Studio Pro rig.

    http://www.gatortraks.com 
    My music.
    ... And of course, the Facebook page. 
    #20
    FastBikerBoy
    Forum Host
    • Total Posts : 11326
    • Joined: 2008/01/25 16:15:36
    • Location: Watton, Norfolk, UK
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/04/04 08:18:06 (permalink)
    I have used quantize but to get the results I want I usually have to select a few transients at a time, and even then I play with the settings a bit.

    It's not as simple as select whole track/clip and quantize to get the results I want but then I find that true of any quantize, be that MIDI or audio.
    #21
    brundlefly
    Max Output Level: 0 dBFS
    • Total Posts : 14250
    • Joined: 2007/09/14 14:57:59
    • Location: Manitou Spgs, Colorado
    • Status: offline
    Re:Audiosnap transient detection is much improved. 2011/04/04 10:29:39 (permalink)
    It's not as simple as select whole track/clip and quantize to get the results I want but then I find that true of any quantize, be that MIDI or audio.

     
    +1 In my experience AS quanize works as well as MIDI quantize with the caveat that you have to make sure the transient markers are accurately placed, and ony the ones you want affected are enabled.
     
    Then, as with MIDI quantizing, you have to make sure your're quantizing to the right resolution with notes close enough to the intended timing that they don't get quantized the wrong direction.
    #22
    Jump to:
    © 2024 APG vNext Commercial Version 5.1