Audio editing and manipulation software

Although I use a tablet computer (and/or CD player) when calling a programme of folk dances to recorded music, I employ a PC operating under MS Windows 7, and recently Windows 10, (64 bit version of both) to prepare dances and programmes, and to manipulate or synthesize music.

For editing or manipulating audio tracks that I have myself recorded of local bands or have synthesised, I use Audacity and/or Goldwave. Audacity has more capabilities and is free software, but Goldwave is quicker and more straightforward for some jobs. To restore recordings that are in a poor state, particularly 78rpm records, I also use ClickRepair and Equalizer.

To illustrate some of the capabilities of Audacity and Goldwave, I will show parts of the processing applied to one of the tracks, the 4th, that we recorded during a practice session of the Kenton Ramblers in 1978. This music is for the dance "Dick's Maggot". The background to this recording session will be found here. The original recording was made on an Akai 4000DS MkII stereo ¼″ tape deck using Agfa-Gevaert PEM368 tape.

The recording has been digitised by a xitel INport USB device with CFB Software LP Recorder. I sample at 44100 samples per second for each of the two (left and right) channels and store the samples as 16 bit integers in an uncompressed format as a ".wav" file.

This Akai deck did not have any tape noise suppression facilities (i.e. no Dolby®). I will demonstrate how tape "hiss" and other extraneous noise can be removed using the noise reduction facility in Goldwave; a very similar capability exists within Audacity. As this was a practice session of the Kenton Ramblers, not all the tracks were played through upto the number of repeats that are required for dancing. A technique for extending the number of repeats is then illustrated using Audacity.


Click picture to magnify 

Audio processing - Goldwave

When the "wav" file for the 4th track that we recorded is read into GoldWave, the initial display shows the time histories of the signals (which were originally air pressures) for the pair of channels: left channel at the top in green, right channel below in red.

All the subsequent processing is applied simultaneously to both channels by the software.

If we zoom-in on the start of the track, we see that there is a brief section of just the noise before the music starts. This is principally 50 Hz mains hum, but there is tape hiss superimposed. These also contaminate the music that begins 1.25 seconds into the recording. Click here to hear the 2.3 seconds shown in screen shot. You will probably have to increase the volume to hear the noise, but don't forget to turn it down again particularly if you are wearing earphones!

To attempt to remove the noise from the recording, we first select a sample of the noise and copy it to the clipboard. The copy operation can be done through the drop-down menus (as shown here), but the keyboard shortcut <ctrl>C is convenient.

Before moving into the noise reduction facility it is vital to extend the selection to encompass the whole track. The quick way to do this is to use the keyboard shortcut <ctrl>A  Without doing this, noise will only be removed from the 0.577 seconds that was selected to characterise the noise; not a very useful operation!

The drop-down menus are then used to reach the "Noise Reduction" control box. Here the only change that has to be made is to select "Use clipboard" from the "Reduction envelope" options before pressing "OK".

Depending on the speed of the processor in your computer, the processing can take some considerable time. Eventually the same section of time history is re-displayed. Click here to hear the 2.3 seconds shown in screen shot. In this case the clean-up is good, but it must be realised that such processing can, and often does, introduce some unwanted artifacts to the sound of the music. You need to listen carefully on reasonable quality headphones to judge the results of this kind of manipulation.

Some of the artifacts, if present, can be audible at the end of the track as the music fades away. Continued "ringing" that dies away in what should be silence is one possible effect that can be introduced. So we move to the end of the track and listen to the results.

There are a couple of other small edits that can be performed.

I prefer to have at least 5 seconds of silence at the end of a track to give time to switch the music off when a dance finishes. You have to listen carefully at high volume to hear when some lingering instruments actually fade out completely so you can be sure that you are not truncating their sound. Then, in Goldwave, select the rest of the track and use the menu item "Edit" → "Mute" to remove all following residual noises. If necessary further silence can be added using "Edit" → "Insert Silence..." or prolonged silence removed by selecting the excess and pressing the Delete key.

I also like to have at least 1 second of silence before the music starts. I find music that starts immeadiately at the beginning of a track, as many commercial tracks do, is difficult when calling as it gives no time to turn round or look at the dancers before the music begins. However, a silence of much more than 2 seconds is confusing as you can think that playback has not started.

So moving back to the beginning of the track, any remaining noises (there is at least one click at 0.64 seconds) can be removed by selecting upto the start of the music and then "Edit" → "Mute". Again silence could be inserted or removed if necessary; it wasn't in this example.

The whole modified track is then written out to a new ".wav" file, that is still uncompressed. Goldwave can write an ".mp3" file, but conversion to such a "lossy" format should be left as a final step once all processing is complete.


Audio editing - Audacity

After completing noise reduction processing, either using Goldwave (as illustrated above) or by using similar facilities in Audacity, we now turn to dealing with another deficiency in the recording of Dick's Maggot. It is only 5×16 bars long, which is a bit short for dancing, while simply repeating the whole dance is likely to be too much. Note that the music is in 3/2 time and so dancers take 3 steps to the bar; once through the dance is 16 bars. We will therefore look at a means to extend the track to 7×16 bars.

Audacity again presents the stereo tracks with the left channel at the top and the right channel below, but they are not coloured differently and the whole display is rather more muted than Goldwave.

Having read the stereo audio into Audacity, we first add a "Label track". The easiest way to do this is to start listening to the music and when the first notable time in the music is reached press <ctrl>M . If there is not already a label track, then one will be inserted and a label placed at the time of the key stroke. The text insertion cursor remains in this newly opened label and we can type a description "1a1" to indicate that this is the start of the first occurence of the A1 four-bar phrase of music. When the repeat of this phrase begins we hit <ctrl>M again and type in "a2".

We continue to insert labels for the start of the "b" and "c" phrases in this tune, and then "2a1" at the start of the second time through the whole 16-bars.

The start of the other repeats are marked as "3a1", "4a1" and "5a1". We then have to decide which 2×16 bars in the recording we wish to be repeated. In this case, I chose those starting at "2a1" and "3a1", so I selected from a few seconds before the start of the "2a1" music all the way upto the end of the track. The selection is copied into the clipboard; this can be done from the drop down menus but the key stroke <ctrl>C is quicker.

We next use Audacity's drop down menus "Tracks"→"Add New"→"Stereo Track"; this facility in Audacity is an extremely valuable capability.

Position the cursor a few seconds before the "4a1" label in this new track, and paste a copy from the clipboard; <ctrl>V is the shotcut key stroke to do this.

Returning to the original upper stereo track, we select the interval from a few seconds after the "4a1" label upto the end of that track, and delete (by pressing the Delete key).

We zoom in to look at an interval of a few seconds surrounding the "4a1" label.

By depressing the "Solo" button at the left hand side of the upper (original) stereo track pair, we can listen to just these, again press <ctrl>M to insert a new label at the note which starts of the "4a1" music, and insert the name "upper" for this label. Had the original "4a1" label been placed accurately enough, this new label should co-incide with it, but such perfection when listening to the whole duration of a track is unlikely. Looking at the audio notes that appear visually above the label track, this new label can, if necessary, be moved slightly to align with the start of the note.

By switching off the Solo button for the original upper track, and instead switching it on for the lower stereo pair, the process can be repeated to insert and position a label "lower" at the start note for the "4a1" music (which was originally "2a1") in this track. Unless you have been very exact (or lucky!) in your selection and placement, these two new labels will not align. There will be a time difference between them.

We now choose a new tool from the toolbar. These are six small square buttons (in two rows of three). Up until now, by default, the I button for performing selection has been depressed. Instead we now depress the left-right arrow button ↔ to perform a time shift. The cursor changes to a left-right arrow ↔. We move this into the lower track and drag the whole of the stereo pair of signals forward by the identified time difference. Ensuring that neither Solo button is depressed (nor the related Mute button), we can then listen to the short interval before and after the "4a1" label. Usually a few more small time shifts to the lower track pair will be needed to get as good an alignment as possible; it is a subjective judgement. The picture on the left shows the final alignment and the result can be heard here.

Remember to change back to the I selection button.

To remove the dupicate instruments that can be heard during the overlap, for the upper track we select and delete from one or two notes after the transition point to the track end. For the lower track we select from the track start upto one or two notes before the transition point. However, we can not just delete this selection as the remainder of this track would move left to fill the gap, so destroying the careful alignment. One way round is to instead silence the selection using  Edit→Remove Audio or Labels→Silence Audio  from Audacity's drop-down menus.

To implement a smooth transition, a cross-fade betwen them is required (but see footnote for changes that have been made in later versions of Audacity). We therefore select, in the upper track only, the interval from the start of the music in the lower track upto the end of the music in the upper track (which should only cover a few notes) and apply  Effect→Cross Fade Out  (not Fade Out which appears first in the drop down menu). Now select in the lower track the same time interval and apply  Effect→Cross Fade In . The use of cross fade out and in ensures that the overall (that is summed) volume of the merged tracks remains constant. The result can be heard here, Having in the past performed this kind of audio editing by physically cutting and splicing ¼″ magnetic tape, I find that the ease and adjustability provided by Audacity is remarkable.

To finish off, we zoom out to view the whole duration of the 7×16 bar track and export this as a ".wav" file. It can also be exported as an ".mp3" file from Audacity, or this convertion can be performed by other software from the ".wav" file.

Before exiting from Audacity, the work can be written out as an Audacity project to an ".aup" file. This maintains the multiple tracks and the label track, information that is not in a ".wav" or ".mp3" file. In fact, it is a very good idea to write a project out soon after you start work on a track in Audacity, and then write out updates as the work progresses so you can restart a project at a later date.

The complete processed and edited track can be heard here Dick's Maggot (7x16).  (file size 4.5 MB, duration 3:10).

There is a video of the Bedford Tuesday Playford group dancing to this extended recording.

Conversion to mp3

For access on the Internet, as given by the above link, the track of Dick's Maggot has been converted to an ".mp3" file at 192 kbps (kilobits per second). This is above the 'default' 128 kbps that is offered by many mp3 conversion software facilities. It gives a file size of about 1/7th that of the uncompressed stereo ".wav" file. For a discussion of the effects on audio quality of (lossy) mp3 compression at various bit rates see the results at mp3-tech.org.

I also use 192 kbps for stereo ".mp3" files that I prepare myself for transfer to my tablet. If I am buying a track or album online, I try to download a version with a bit rate above 128 kbps for an ".mp3" file.



† The steps needed to introduce a cross-fade have been changed in more recent versions of Audacity. Applying the new method is rather different and the (stereo) track to be faded-out must be immediately above the track that is to be faded-in (which happens to be true in the current example).