audacious-appendment

An Audacity script to merge a series of mp3 tracks into one, and perform some minor effects.

Usage

Say you have a series of tracks, e.g. the three movements of a concerto. You want to merge them into one track.

Run the script on the three movements, and if you want, truncate the silence of their tracks (to be fully implemented); and if you want, determine where to add silence and how much to add. The script will do it all in Audacity, using Audacity commands and a few Nyquist commands (no worries, no plug-ins required, luckily). (Sorting hasn't been implemented yet, so the order of the tracks should rely on the order of arguments passed in.)

./script.py "medtner 1.mp3" "medtner 2.mp3" "medtner 3.mp3" -o mechamedtner.mp3

Functionality

The script is very basic at present. It should:

1. Read files pass as arguments, determine their path, and create a .lof file from them
2. Open Audacity, and import the .lof file
3. Merge the tracks together, end to end
4. Add some effects, such as amplification or truncate-silence
5. Export the result to a path

Future

There is much to be added. I want to add a sort option for files that are named in a sortable way (classical music). I want greater options for adding silence to the start/end of tracks. I want to make the script nicer. I want to add incremental track addition, so that I can merge more than a dozen tracks together (Audacity's track limit is 16; if I want to combine several tracks a la some Vikingur Olafsson album, then I need to implement this). I need to add Linux/Mac support, for no reason but robustness. Also, the script doesn't work if this instance of Audacity's pipes had already been used earlier.

Audacity's capabilities are also limited. If we could specify durations on commands, it would be great and extremely convenient. If we could control export options, it would be extremely great. If Python could have a better argparse module, that would be excellent.

Nyquist Commands

The Nyquist-LiSP commands were challenging to figure out. The documentation is austere; the community is absent; the symbol names are unintuitive.

And of course, s-expressions are tangly, if you haven't written LiSP or read the revered book on the structure and interpretation... (But hey, s-expr parsers are so easy to write, aren't they?)

The capabilities of Audacity's scripting API is itself still somewhat immature, but luckily it turned out to have everything I needed to write what I wanted. There were a lot of times I had thought I was stuck - for example, I relied on the Generate: Silence effect in the GUI, but it was not available in the scripting API. I eventually decided to just make noise with an amplitude of 0. But while this would work in the GUI, I found out I couldn't specify a duration from the script! - and so a brief quest to study elementary Nyquist at 3:30 AM kicked off. And even when I found something that worked when I entered it into the Nyquist prompt in the GUI, it didn't work in the script! Turns out it wouldn't work for multiline strings; so I took advantage of s-expr's lack of needed whitespace and put it all in a one-line string.

; Example of a Nyquist comamnd that inserts silence at the beginning of a selection.
(defun insertstart (sig) 
  (sum (s-rest 2) 
    (at 1 (cue sig)))) 
(multichan-expand #'insertstart s)

Audacity defines multichan-expand for use with stereo channels, since Nyquist can only apply functions to one channel at a time (and represents stereo as arrays of channels).

Actually, to make this add silence to the end of a selection instead, it's as easy as changing the (at 1) to (at 0).

Overall, you can imagine this command to say something like: define a function "insertstart" that takes a signal (a sound) and gives the sum of some silence and the signal but shifted over by its duration. Then call the function on both channels. Gosh, Lisp is cool and ugly at the same time. I love ML; but I respect LiSP all the same.

Hopefully, if anyone ever has the problem of inserting silence at the beginning and end of a file, they can look at my script and figure it out themselves.

Credits

Thanks to Steve Daulton for his Nyquist plugins, which I pretty much copied here.

Thanks to the Audacity team for their helpful pipe_test.py and pipeclient.py, the latter of which I more or less used to not only test my script but scaffold it. I plan to make this script single-threaded (use no actual threading at all) later on, as it is a blocking script anyway; that will release any old code from the pipeclient.py the team wrote.