How do I append a bunch of .wav files while retaining (not-zero-padded) numeric ordering?

Question

I have many (7500) .wav files (named by their numbers) which I'd like to merge into one. I think it has to be done with a bash loop because with using globbing some files would be in the wrong order (like 1111 before 2).

What's the best way to do this under Linux?

Solution

If all the files have the same parameters, like sample rate and number of channels, you still can't just catenate them. You have to strip away the WAV header.

It may be easiest to use an audio file manipulation utility like sox, which contains a method for catenating files. In fact, it does this by default. E.g. to combine three .wav files into one long one:

$ sox short1.wav short2.wav short3.wav long.wav

A loop will not arrange the files in the order you want. What you want is to sort the names, but treat them numerically. sort -n will do this.

Thus:

$ sox $(ls *.wav | sort -n) out.wav

If sox cannot handle that many files, we can break up the job like this:

$ ls *.wav | sort -n > script

Then we have a script file which looks like:

1.wav
2.wav
...
3999.wav
4000.wav
...
file7500.wav

We edit that to make several command lines:

# catenate about half the files to temp1.wav   
sox \
1.wav \
2.wav \
... \
3999.wav \
temp1.wav

# catenate the remainder to temp2.wav  
sox \
4000.wav \
... \
7500.wav \
temp2.wav

# catenate the two halves
sox temp1.wav temp2.wav out.wav ; rm temp1.wav temp2.wav

As the first editing step on the file list, you can use the vi command :%s/.*/& \\/ to add a backslash after every line.

Comments

[Neula]

Thanks,this seems to be the right way, but it seems like sox doesn't take more than 6479.wav files (quits with "sox FAIL formats: can't open input file `6479.wav': Too many open files") :-(

[Kaz]

Ah, in that case what we can do is simply break the operation down. For instance, we can divide the list of files into two halves, merge each half and then merge the halves. I would capture the output of ls *.wav | sort -n into a file, turn the list into sox command lines and then source it.

[Kaz]

Keeping all the files open simultaneously looks like a bug/deficiency in sox.

[Neula]

In the end it showed up that sox doesnt'take more than 348 files so I would have to do this 21 times. When merging the first files I noticed that they seem to be damaged, so merging them doesn't help me. Anyway, you solved the problem! thank you

[Kaz]

You mean the original files are damaged? That's too bad.

[LonnieBest]

Be sure to put your desired output-file-name on the end of that argument list or you'll overwrite the last file listed! Luckily, I had another copy. Pressing enter too soon with sox is dangerous.