About a week ago, celenus turned me on to the phenomenon known as Dictionaraoke. That site defines the musical genre as "audio clips from online dictionaries sing the hits of yesterday and today. The fun of karaoke meets the word power of the dictionary." Karaoke and dictionaries are two of my favorite things, so I figured I'd have to try my hand at making one.

I decided on the song "Can't Fight This Feeling" by REO Speedwagon, because not only is it a sweet monster ballad, but it was the last karaoke song I sung, and I felt like I could have done a better job at it. And the whole reason I picked that song to begin with is that I heard it in the supermarket earlier in the day, and it struck me as a great karaoke song. This whole situation seemed so cosmic, and making the song into dictionaraoke felt fitting.

First, I wrote a script to case-and-punctuation-insensitively¹ remove duplicate words from the lyrics (which I'll leave as an exercise to the reader, I just set each word as a key in a perl hash).

Next, I took the output and ran it through a script to grab the associated sound sample from Merriam-Webster Online.

Almost every file is in the form:

http://cougar.eb.com/sound/[first_letter_of_word]/[first_7_chars_of_word]1.wav

Words shorter than 7 chars are padded with zeros, so every file name is always 8 chars. Eg:

http://cougar.eb.com/sound/f/find0001.wav

The following block of perl got me most of the way there

foreach my $word (sort { $a cmp $b } keys %hash) {
    my $len = length($word);
    my $cmd = "wget http://cougar.eb.com/sound/" . substr($word, 0, 1) . "/"
        . ($len > 7 ? substr($word, 0, 7) : $word)
        . ($len < 7 ? "0" x (7 - $len) : "")
        . "1.wav -O " . $word . ".wav";
    `$cmd`;
}

Don't ask me why I sorted the lyrics.

It's probably against m-w's TOS, and it definitely pegs their servers for a minute or so depending on how big your song is, but it beats looking each word up.

This script got me most of the words. Contractions and inflections have different file name rules, so I just grabbed them by hand.

Next, I found an instrumental (read: midi) version of the song, and placed the speech samples in the right positions (the most time consuming part). Here's the finished product:

Can't Fight This Feeling - REO Speedwagon [6mb .mp3]

---

¹ Since punctuation was in my input string, "You" and "You," would be considered unique words. Only now did it occur to me that it would have been easier to strip out non alpha characters.