Timing audio - where am I going wrong?

[Artspoke]

westla,

I'm really excited that you tried the experiment, and it seems to have worked for you! Maybe we can make sure we've done the same thing, and if so, maybe we can find out what's different between our setups. Something is different for sure.

I'm going to take a few screenshots and post them tonight. I really hope we can discover the difference between your system and mine, or your process and mine in order to find this problem.

I have been leaning away from the operator error theory due to the fact that the three macs I own, plus three others I use regularly at local studios ALL do exactly the same thing. Also, there are many threads around reporting the same behavior.

Just to verify quickly - you used command plus the sideways hand to stretch a soundbite, and when you finished, the soundbite was exactly one beat long? Further, the left mono track, right mono track, and both sides of the stereo track are all the same length to each other, and there is no mystery pop at the end of the new files?

What is so exciting, is that for the first time, we've got two different outcomes to the same experiment. If the experiment is at fault, we'll find it, and if there is a solution, I think we'll find that too!

Thank you for checking this out, and believing there is at least a reason to have a look.

Seeing that you might have a working system, whereas there could be a flaw in all of the systems I've tried, would you mind trying a couple other things? I have been fighting with this for quite awhile, wihch has led to many theories as to where this problem is, and what it affects. If I were to post a small session someplace, or have one available to email, would you or anyone be interested in trying that?

It's seemingly a small detail to those who don't see it, but if you have it like I have, you definitely can't stand it!

Thanks for getting this rolling a bit at last, I'll report back. Anyone else have this WORKING? Actually, I think anyone who could post results to this little stretching exercise would be helping immensely.

Thanks some more,
-James

[musicarteca]

I also tried the experiment with two small variations: I used a MIDI source for the click instead of the one generated by the computer, and I copied the soundbites, pasted them at the new location and then changed the tempo, it seemed more of a logical procedure that the copy, change tempo and paste that you suggested.

Results: Visuals: nothing strange whatsoever except for the fact that the mono tracks were recorded with a little bit more volume, and showed a slightly larger release tail than the stereo track. Sound: No artifacts of any kind whatsoever, the sound is clean in every track. No clicks or bumps at the end.

I don't think that I have much more to add to this thread, I have written everything that I wanted to say. Some others are welcome to pitch in.

<small>[ April 16, 2005, 07:06 AM: Message edited by: musicarteca ]</small>

[slybass3000]

Hi James!
One thing I learned here in this forum is the fact that the Internal click of DP is very bad. It is in relation of your buffer size. If you have a high setting like 512 or even 1024 , you'll get a lot of latency while you listen to it and when you record it. MIDI click is a bit better and the best is Hiro's clippings clicks for DP4. This is totally accurate.
And for the artifacts you get, may be it is coming from your Clock. What kind of Clock are you using? This might be the problem.
Hope this will help,
SB

[Artspoke]

Interesting idea about the clock. I'm using the Internal clock on my pci-324 card. I have 3 2408 mkII's hooked up to the card, but I don't notice any clicking/etc. = clock related problems. I did try clocking the system to an external SPDIF clock, but experience the same issue.

The recording of the click I've made looks and plays great. In fact, this is true of all the recordings I've made with this system. The problem I'm having is with stretching audio (one of the problems anyway).

I believe there is a solution to my problem, solely due to the fact that Alex Rodriguez has his system performing this without the glitch that mine is. This gives me hope.

It is frustrating when someone doesn't just fly to the rescue with a simple setting change that fixes your entire problem, however, it is important to realize that all experience related to your problem is useful. The experience of the other users on this forum is invaluable - whether it's the same experience or not!

I believe there is something that differs, perhaps at the system level between my, and say, Alex's system, that is causing DP to behave in different ways. I have performed a clean reinstall of Panther, and then updated it via the Apple site to 10.3.8 (soon to try .9). I then reinstalled DP 4.5 and updated to 4.52.

At this point, before installing NI, UAD-1 (sortware and hardware) or really anything else at all, I launched DP at last, and tried to stretch a soundbite. Sadly, I have the same problem as described before.

I would consider just buying a new computer altogether, except that I've tried this on other computers and the same thing happens.

If it were not the case that I've read and posted to other threads started by users experiencing the same or similar problems, I'd just give up and check myself into the looney bin!

Resonant Alien, Timeline, and the others who have posted regarding time scaling wierdness, if you have found solutions, I'd love to hear them.

Meantime, someone posted a snippet from an email from MOTU regarding this issue. Their response was that PureDSP pitch adjustment could be a factor, as there are known problems with it. Does anyone have any information about this? Does it in fact have any effect time scaling? What are the issues?

The work I do on film, not to mention straight up music production, relies on this technology. It is keeping me alive knowing that some users are reporting trouble free operation. What could be possible reasons that it works for some and not others? I mean, where, specifically, might a difference between systems cause behavioral differences, system to system?

Alex, thank you for your time and energy regarding this thread, I'll miss your input here. If/when I do find a solution, I'll be sure to post it to this thread, so I guess it's safe to say that I am not finished posting to this thread - I do have more to say, and hopefully, more to read.

Thanks again,
-James

[Artspoke]

At last, I have a solution.

The analysis files are the key to the solution - indicating that the analysis itself is on a track by track basis, and there's nothing you can do to two tracks to get them to analyze the same way, in terms of tempo, for the purpose of tempo-based stretching. Copy Beats does not work. Copying the tempo from the same sequence doesn't work either, even recording the tracks in the same sequence does not allow the Adjust Soundbites to Sequence command to stretch all of the recordings by the same amounts, beat per beat, as is necessary to retain phase coherency between the mics.

Ok! To get stretched, yet phase-aligned drum tracks, from the top:

1) Record the kit with several mics

2) Project -> Modify Conductor Track -> Adjust Beats. Zoom up and drag every (conductor track) beat so that it matches the drummer. When this is complete, the metronome will follow the drummer exactly.

3) ***New steps*** Create an additional mono audio track. Copy a single snare shot from your snare track, and paste it on quarter notes in the new audio track, creating, in effect, an audio click track.

4) Select exactly from the beginning of any of your drum soundbites to exactly the end of the soundbite (I like to use the tab, shift tab method). Now click on the new snare/click track name, trasferring the selection bounds to this track, and Merge the audio (make sure there's some audio at the end of your selection, the goal here is to create a new soundbite that's exactly the same length as each of the drum soundbites).

As this newly merged soundbite was created under the tempo map based on the drummer, it will automatically contain this tempo data.

5) Create a new take on the conductor track, and change the tempo to your goal tempo - perhaps the tempo that the click was playing into the drummer's headphones.

6) Select the newly merged snare/click soundbite and Audio -> Adjust Soundbites to Sequence Tempo. This will work beautifully, so now the snare/click soundbite that used to follow your drummer will be "fixed" to the new tempo.

7) Select the first soundbite in your kit - probably the Kick and Audio -> Adjust Soundbites to Sequence Tempo.

Hit Undo!

9) Hide DP and go look in your project folder. In the Analysis Files folder, there will be two new entries - one for the click/snare, and another for the Kick (ignore any zero kb files that show up).

10) Highlight the name of the Kick file, copy the name, and then delete this file.

11) Right-click the snare/click analysis file and select Duplicate

12) Rename this file by pasting what's in the text buffer (in this case, Kick)

13) Go back to DP, select the Kick soundbite again, and Audio -> Adjust Soundbites to Sequence Tempo. This time, you'll notice that Background Processing doesn't re-analyze the audio - it goes straight into Adjusting Tempo. If you do see it analyzing the kick - something didn't work. If it analyzes the soundbite at this stage, it will alter the analysis file, and the stretch will be corrupt.

Repeat 7-13 for each soundbite in the kit, making sure to always duplicate the snare/click analysis file for each soundbite. Once you figure out the name convention, you don't have to keep repeating steps 7-10, but it does help keep track of what file you're on. I find it useful to sort the contents of the Analysis Files folder by Time Created.

Ok, so this is completely crazy! But, it works, and as far as I know, there is no other software that will allow this level/style of audio manipulation. What I'd love is a simple "Use analysis file from:" pulldown someplace in a future DP release.

You won't believe what you hear when you get all the mics stretched - it's artifact free, and phase coherent, and I mean it. I'll bet, at the time of writing this, I am the only human being ever to hear drums straightened in this fashion. You could be next!

[kassonica]

WOW I bet the groove is err.....................

Never mind

Good technique though. Thanks for the post

[Spikey Horse]

........ I'll bet, at the time of writing this, I am the only human being ever to hear drums straightened in this fashion. ...

Actually, I think I saw it done on an episode of Dr Who* once

.. Interesting thread - I think I need to sit down and read it all the way through though.....

* oh wait he's not human strictly speaking.

[Artspoke]

The interesting thing is, the groove is potentially stronger, in that, the drummer's intentions are revealed. It's up to you how close you want to cut. If you choose quarter notes, the stretching merely takes the beats and puts them, well, on the beat. The eighth,s and swing, and etc. are between them. If you were to choose downbeats of every measure, or every four measures, you'd still cure a lot of drift, without removing any feel at all. Maybe you just want the whole song to be a slightly different overall tempo, and not affect the feel at all. You can't do that any other way, so far as I can see. Try stretching a bunch of drumset mics - use the grid, or groove quantize, or quantize beats within soundbites - whatever you choose, you will have an unruly, unusable outcome - it doesn't even work a little bit.

The point here is and has been, for nearly five years since this thread started, the claim that DP can stretch/quantize audio, across multiple tracks, and retain phase coherency, has at last been shown, in this thread, to be true. I've taken some heat over this topic for "over-teching" what would normally be a more musical experience, but threads about musicianship belong in player forums, not here. Should you need the ability, in whatever circumstance or amount, to perform this modern miracle, you now can, with DP, and perhaps nothing else. The ability to autostretch, beat by beat, from one tempo to another, was heretofore simply not offered.

I'd love to get someone else to get this to work, and then start a new thread about exciting, undiscovered uses for it. Any takers? I had trouble getting DP to believe the initial analysis file's clones were kosher. If there's anyone out there that knows anything at all about what these analysis files contain, and their conditions, I'd like nothing more than to learn. Escept to be able to open and edit them!

Of course, another option would be for MOTU to FIX THE TIME STRETCH BUGS, which would make this functionality instantaneous. That I can do it with this silly analysis file swap trick means that the software can do it - one of the bugs in the stretcher has to do with this very problem. If it were possible to simply tell performer to stretch the audio without first "analyzing" it, and base the stretch 100% on the tempo data or the analysis file of another file, that bug would squash. A lot like Groove Quantize, only functioning. The other time stretch bugs are perhaps slightly outside the scope of this thread, but if you've ever messed around with the time stretching tools in DP, you've undoubtedly noticed manifold mayhem in the forms of artifacts etc.

Nutshell question: Why doesn't performer have the ability to stretch audio based solely on tempo information? With the Record/Adjust Beats and Copy Sequence Tempo to Soundbite commands, you can do a stellar job of communicating the timing of a musical recording. Adding the Adjust Soundbites to Sequence Tempo functionality, MOTU has implied that the user actually has these abilities. Which is why I'm posting, to let you know that it is indeed possible, it's just a wonky pain in the behind.

Were I a computer programmer, employed by MOTU, I should be proud of the simple, ingenious solution I would work hard to provide st least this issue. Were I in marketing, I would steer away from truthfully suggesting that the time stretching features in DP are unacceptably awful.

...this post was longer! Admission: I'm and editing nutjob! Well at least sometimes,
-James

[bongo_x]

Wouldn't it be easier and more predictable to just manually cut and move the drums to where you want them? It seems to me people spend a lot of time and energy trying to get "automatic" things to work.

bb

[rainmaker]

I agree. For me, the quickest way to slice up a drum track and have it be perfectly on time and phase coherent is to group all drum tracks together, then zoom in and start splitting the soundbite at every beat with "Apple-Y". Once finished, just select all the drum tracks and quantize to whatever note value you want. Then just a quick listen to crossfade any spots that created gaps. I can do an average length drum track in about an hour and the end result is perfect since the human touch was there all the way for every beat.

Thanks,
Rainmaker

[chamelion]

Being the kind of guy who needs technical frustrations like a hole in the head, I invariably look for the line of least resistance. If and when I have a problem getting DP to accurately process a drum or percussion track as per the above example, I almost always fire up Recycle and convert the audio into a Rex file. Just drag the Rex file into DP, quantize the slices, and everything lines up perfectly. Phatmatic Pro does essentially the same thing. Works for me.

Cheers,

Geoff