Topic: Audacity (osx): How to mix two AUD sources?

[TheBishop]

Thanks, I appreciate that. Next up is a Toubab Krewe show that I've been needing to master for several months now.

[morst]

Bumping this thread, because I just re-wrote the instructions for my morst-Audio-Speed-Converter.XLSX spreadsheet, which is available for free download still at this link:
https://docs.google.com/spreadsheets/d/1pQGfYwPgBFFzcY5m6aRj-Zbu9HsRumLy-tJB1d8Eufg/edit#gid=583050244

"This spreadsheet is a tool which can calculate the percent of ""Speed Change"" needed, in order to synchronize two uninterrupted audio tracks, made with different digital recorders which did not have timing clocks locked together.
This tool was made for use along with the Audacity workstation software. Audacity is availble as a free download for PC, MAC, and Linux. This tool's results can likely be used for other workstation software, as well."   


"Before using this tool, the user should precisely locate and mark four points within the two tracks to be synchronized, in order to make find the two values to feed this tool.
To make a track mark (label) in Audacity on PC, use the mouse and click the point in the track, then use Control-B keyboard shortcut, or go to Tracks > Add Label At Selection. On MacOS, it's Command B.
For greatest accuracy over long files, the two points at which the files are be synchronized should be as far apart in time as is practical, that is, one near the start of the shortest segment, and one near the end of it.
As each point is located and marked, the start time of the file to be changed should also be marked.
The two marked start times are where the difference in time between the sync points of the file to be changed may be most easily measured, as the waveform of the first spot moves from its marked location when the user finds the second spot.
Rather than having to find the same bit of the pattern to mark again, it is much easier to simply use the start of the track as a proxy.
If it is not practical to use the start times, the file end points may alternately be used."   
                                       
"To use this tool, the user enters two values, and the desired output is the Speed Change value (% differnce), which can then be used in the AUDACITY workstation software's Speed Change plugin, in order to synchronize the tracks.
The primary version on lines 4-5 allows the user to enter the length of the reference file, in samples, and the number of samples in between the marked start points. (see above)
The optional version on lines 9-10 allows the user to enter two track lengths, in samples. This might be useful for matching a soundtrack to the scratch audio of a video production, for instance, or making guesstimations"                                                                       

comments welcome?!

[nak700s]

I'm having a hell of a time right now...

The venue that I have been recording at has given me access to their soundboards so that I can mix them with my audience recordings.  Awesome!  So I take the SBD (recorded at 48k) and my AUD (also recorded at 48k) and load them in to Audacity v2.1 (I was using the newer version until I read this thread so I uninstalled that and installed 2.1).  I lined both of the recordings up at the 0:00:03.725 (178,791 samples) point and then find a common point towards the end of the show, for the soundboard this is 2:12:43.275 (382,237,217 samples) and for the audience this is 2:13:05.196 (383,289,428 samples).  I use morst's Audio Speed Converter spreadsheet:



and I get a % speed change of -0.27527696236864200000.  I put this in to Audacity (without the negative sign) and run the Change Speed effect on the audience recording:


This is where things get funky...

At 0:00:27 the recordings are off by 77 msecs
At 2:15:00 the recordings are off by 500 msecs
but at 1:08:00 the recordings are off by 6 seconds!

This is my first time trying to combine two sources and I'm going nuts.  I don't understand how they could be so close at either end of the show, but so far apart in the middle.

If this is something you plan on doing more often, may I suggest an equipment upgrade...  You may want to consider a 4-track deck that you can record 4 tracks internally.  I use the Sound Devices 744T, and there are several other great options as well, like their Mix-Pre series.  A long set of cables and you're now simultaneously recording all 4 tracks (a pair of mics + a stereo soundboard feed) on the 1 deck.  All four tracks will open in your editing program, and you'll decide what to bring up or down to your liking.  It's about 1000% easier!   

[checht]

If this is something you plan on doing more often, may I suggest an equipment upgrade...  You may want to consider a 4-track deck that you can record 4 tracks internally.  I use the Sound Devices 744T, and there are several other great options as well, like their Mix-Pre series.  A long set of cables and you're now simultaneously recording all 4 tracks (a pair of mics + a stereo soundboard feed) on the 1 deck.  All four tracks will open in your editing program, and you'll decide what to bring up or down to your liking.  It's about 1000% easier!   

^ this. 100% agree.
Went from Mix-Pre 3 to 6 for this reason. Had an old pair of mogami 100' cables, added some 50's & 20's from Tgadakis and everything got better. Next we can discuss the ultimate SBD connector field kit 😀.

[morst]

You can't always get a cable between the SBD and the best sounding spot, so this technique won't go away just yet.

OH FUCK this shit is all old and probably solved.
nevermind

That correction number is VERY large. Are you sure the stretch/squash requires six seconds of correction?
1052211 is a lot of samples...
Are you certain that both recordings are continuous? No pauses?
I think you should consider using the other set of lines in my spreadsheet and entering the length of the longer section as 382,058,426 and the difference as the number of samples between the early Sync track start point and the later Sync track start point
When you synced at the start and the end, please mark the locations of the file starts, and use the amount of change as the DIFFERENCE.
The length of time BETWEEN the points is only 382,058,426 samples, which I suggest to plug in as the "LONGER" Length.
(I am gonna have to rewrite this spreadsheet soon anyhow so it's good to get clarity/communication issues sorted out.)


** A note for clarity - the speed change is not computed based on the length of the entire fire.
The speed change is computed based on the number of samples between the two sync points on the longer source versus the number of samples between the two sync points on the shorter source.
I find it easier to count the long one and mark the shorter one's movement, which is why I use the set of lines in my sheet where I plug in length and difference.

[justink]

Let me give it a try for ya. These instructions should work fine on a mac or a PC running Audacity. Probably Linux too, though I have never tried it.

1- Run the masters so the levels peak below -3.0 dB on each if you can, so you can mix them without having to lower volume to avoid clipping peaks. Run them both at the same sample rate. Bit depth is not as important to match. Name each file descriptively including the type of source (Foobar2008-04-29-SBD-24bit.wav, Foobar2008-04-29-DPA-24bit.wav or some such)

2- Go to AUDACITY>PREFERENCES> QUALITY and set your default sample rate and bit depth to the desired settings (44100 with the desired output bit-depth). PROJECT>IMPORT both sources into audacity and FILE>SAVE your work.

3- Use the double-headed arrow "TIME SHIFT" tool in the top left corner (<-->) to line up the files as near as possible to the start of the music. Find a sharp peaking impulse like a drum hit or some other peak to line up and zoom in until they are as precisely aligned as possible.

4- If you think you need to do any EQ or level changes on either source, do this now. You can use EFFECT>AMPLIFY to tell you how high each source peaks. If each source peaks higher than -3.0dB at any point, you will get clipping on the mix, so you'll need to lower levels to avoid this. It is also possible to have one source higher than the other and still clip peaks. I am good but not great at math, so hopefully some other folks can chime in with suggestions and comments regarding levels. I think I have it right, that 2 signals of -3.0dB will add up to peak at -0.0dB so you want to stay below that.

5- Go to the end of the files and figure out how much they have drifted apart. Do some math* to figure out how much you need to use EFFECT>CHANGE SPEED to get them lined up (see below for my method of calculating the percent change)

6- I suggest shortening the longer file rather than stretching the shorter one, but it probably doesn't matter. Use the EFFECT>CHANGE SPEED to do that.

7- Check the alignment to make sure the sources stay together. If they are not correct, use EDIT> UNDO SPEED CHANGE and try step 6 again. When they are correct, FILE>SAVE your work again. As long as you keep the file open, you can UNDO past the file save operation, but once you close and open it again, you can't go back past the saved version.

8- Check the mix for sound by using the MUTE function on each track during playback to make sure it sounds good. Adjust the gain for each track if needed by using the +......- slider on the left of each track for course adjustment, or EFFECT>AMPLIFY for finer control.

9- Go to AUDACITY>PREFERENCES > QUALITY > HIGH-QUALITY DITHER and select "Triangle Dither"

10- FILE>EXPORT AS WAV (or AIFF) to make the mixed file. FILE>SAVE again. If you think you might need to make further adjustments after checking the completed file, keep this project session open so you can UNDO back. Name the file something descriptive like Foobar2008-04-29-MIX-24bit or Foobar2008-04-29-MATRIX-16bit so you can distinguish it from each source file.

11- FILE>OPEN a new Audacity project document and PROJECT>IMPORT the newly created mix file. EFFECT>AMPLIFY to check that all peaks are below -0.0dB. If this plugin does not offer to boost levels, then you probably have clipped a peak somewhere, and you will want to go back to the original files and lower the levels of one or both sources to preserve your dynamics and avoid flattening out peaks. If you have a little headroom and it sounds good, then you have successfully mixed your sources.

12- If you want to track for CD's, then VIEW>SET SELECTION FORMAT > CDDA min:sec:frames 75fps and then EDIT>SNAP TO> SNAP ON to allow you to cut tracks without "sector boundary errors." Select tracks in order by using EDIT> MOVE CURSOR TO TRACK START (I go into preferences and give it a keyboard shortcut to make this easier) then shift-clicking on the end of each track, then EDIT> SPLIT each track apart in order, making sure to split the final track too.

13- FILE> EXPORT MULTIPLE (NUMBERING CONSECUTIVELY) to WAV (or aiff) in your selected target directory.

14- Compress these files losslessly using your favorite FLAC encoder, and upload to your favorite sharing website, and post in the KICKDOWNS thread here so we can check it out.

15- please let me know if this is unclear or can be improved upon. 


* oh shoot, now I gotta figure out how to tell you the math part! My apologies for the half-assed nature of this part of my method.  Go to VIEW>SET SELECTION FORMAT > SAMPLES (SNAP TO SAMPLES) so you can measure the length of your program in samples. Measure the total length from your sync points early in the file to the desired sync points late in the file. You will get different numbers for each file since they are probably not lined up perfectly due to slight variances in the clock chips of the two recorders. Make a note of each of these numbers. Subtract one from the other to find out the number of samples of drift at the end, and write this number down. Divide the length of the longer one by the length of the shorter source and you will get a number close to but greater than 1.0000000. Let's use an example where you have exactly one second of drift at the end of exactly one hour at 48KHz. The longer file is now 172,848,000 samples and the shorter one is 172,800,000 samples. Divide the long one by the short one and you will get 1.0002788. (If I am getting this right, then) this tells you that you that you need to speed change the longer file by -.02788%

Damn I hope I got that right. Please won't someone troubleshoot my math and let me know the best way to do this??? 

this is not working on new audacity.  doesn't seem to want to take negative numbers.  i'm not having a math issue, it's an issue that audacity either deleted that feature, or changed the "change speed" to not allow negative numbers.  i tried to do the opposite and stretch the smaller file but that didn't work either.

You'd think you could just highlight from the start peaks/sync to the end peaks/sync and drag and stretch it.... but no...

anybody have a new solution or workaround?

[AbbyTaper]

Speed change the shorter one by the inverse.  Although in my opinion the "math method" isn't the best way of matching the sources.

[justink]

Speed change the shorter one by the inverse.  Although in my opinion the "math method" isn't the best way of matching the sources.

I tried that. audacity wouldn't accept it.

IDK what I did, and IDK if I can replicate it again but I figured a work around.

I did something like measure the samples from the "end" peaks.  Got that small sample number.  Then I trimmed the longer file down/deleted the rest.  Then I used "speed change" to match the samples of the longer one to the shorter one (didn't use the speed multiplier).

I honestly don't know how I came to this decision, but I tried it and it worked.  I'll give it a go again next time. 

I just wish it were easier, drag and stretch, or something...  This shouldn't be this hard with all we have now.  lol

[AbbyTaper]

The trouble with the math method is that it assumes the drift is at a constant rate.  In my experience this is never the case.

[nulldogmas]

The trouble with the math method is that it assumes the drift is at a constant rate.  In my experience this is never the case.

With two digital sources, really?

I've never had the math method fail me unless there was a hiccup with one recorder that caused a skip in time.

[AbbyTaper]

The trouble with the math method is that it assumes the drift is at a constant rate.  In my experience this is never the case.

With two digital sources, really?

I've never had the math method fail me unless there was a hiccup with one recorder that caused a skip in time.

Every time, to the point I don't waste my time with the math anymore.  And I was a math major, those many years ago

[morst]

The trouble with the math method is that it assumes the drift is at a constant rate.  In my experience this is never the case.

Depending on your precision needs, of course it will never ever be correct due to fluctuations in air temperature and pressure, so this whole thing is a fool's errand.


Or, it will be fine.


Always works fine for me when sources are free of discontinuity and the mics do not move relative to the PA system.
Hat-worn is a crap shoot since distance varies constantly.

[morst]

this is not working on new audacity.  doesn't seem to want to take negative numbers.  i'm not having a math issue, it's an issue that audacity either deleted that feature, or changed the "change speed" to not allow negative numbers.  i tried to do the opposite and stretch the smaller file but that didn't work either.

You'd think you could just highlight from the start peaks/sync to the end peaks/sync and drag and stretch it.... but no...

anybody have a new solution or workaround?

Oh yeah audacity changed everything at least once.
See my current calculator spreadsheet which allows you to figure the new total length of the section you want to change, and enter the precise number that way.
https://docs.google.com/spreadsheets/d/1pQGfYwPgBFFzcY5m6aRj-Zbu9HsRumLy-tJB1d8Eufg/edit?usp=sharing


I use audacity 2.4.2 because anything newer than that "phones home" (no thanks)

[rocksuitcase]

this is not working on new audacity.  doesn't seem to want to take negative numbers.  i'm not having a math issue, it's an issue that audacity either deleted that feature, or changed the "change speed" to not allow negative numbers.  i tried to do the opposite and stretch the smaller file but that didn't work either.

You'd think you could just highlight from the start peaks/sync to the end peaks/sync and drag and stretch it.... but no...

anybody have a new solution or workaround?

Oh yeah audacity changed everything at least once.
See my current calculator spreadsheet which allows you to figure the new total length of the section you want to change, and enter the precise number that way.
https://docs.google.com/spreadsheets/d/1pQGfYwPgBFFzcY5m6aRj-Zbu9HsRumLy-tJB1d8Eufg/edit?usp=sharing


I use audacity 2.4.2 because anything newer than that "phones home" (no thanks)

+T   I keep the installer for 2.4.2 on hand for when about every two weeks 3.5.xx installs itself on my Windows 10.

[justink]

this is not working on new audacity.  doesn't seem to want to take negative numbers.  i'm not having a math issue, it's an issue that audacity either deleted that feature, or changed the "change speed" to not allow negative numbers.  i tried to do the opposite and stretch the smaller file but that didn't work either.

You'd think you could just highlight from the start peaks/sync to the end peaks/sync and drag and stretch it.... but no...

anybody have a new solution or workaround?

Oh yeah audacity changed everything at least once.
See my current calculator spreadsheet which allows you to figure the new total length of the section you want to change, and enter the precise number that way.
https://docs.google.com/spreadsheets/d/1pQGfYwPgBFFzcY5m6aRj-Zbu9HsRumLy-tJB1d8Eufg/edit?usp=sharing


I use audacity 2.4.2 because anything newer than that "phones home" (no thanks)

I have three versions installed but they're all 3.x.x plus.

I tried to install 2.4.2 but it said that my OS was not cooperating with that version.  :shrug