In my travels, I found that people use two different models for creating music/light shows. I'll call one model "scripted", and the other "interpreted".
The interpreted modelThe interpreted model is easier to set up quickly, but the results don't necessarily look as good. In this model, the software listens to various properties of the music as it plays and interprets what channels on the relay board to activate. This is the same principal used by the visualizers on your favorite media player software. Think of this model as the plug-n-play solution. The same software will play the music and send data out the parallel port.
The results won't look horrible, but it really depends on the song. I used "This is Halloween" from the film "The Nightmare Before Christmas" in my example movie. It came out pretty well. Some other songs don't work as well. On some songs, the lights didn't really blink that often so the illusion of singing was lost. You may be able to mitigate this issue by configuring your software better.
The scripted modelIn the scripted model, the task of determining what channels to activate is done by a human. You use a piece of software that will play the music to you while you tap on your keyboard when you want to activate a channel. The software will record your tapping. This can be time consuming since you will have to listen to a song multiple times to tap out the sequences for each of the 8 channels. You will use the same software to play back the music to your speakers and the data to the relay board. If you're good, you can get pretty decent results using this method.
The hybrid modelThere is actually one more model. It's actually a mix of the two. It involves recording your own vocal track of a song. Then you use software to write a script by interpreting your vocal track. A vocal only track is used so that the software doesn't interpret any instruments in the music as voices. The illusion of the singing pumpkins is increased because the pumpkins will only flash to the vocal track. I read about this model at HalloweenForum.com
What did I do?I decided to go with the interpreted model. I did this for a couple or reasons in no particular order: laziness, time constraints, it looked good enough to me, we could have a wider selection of songs more easily.
For software, I used Winamp combined with the DiscoLitez plugin. The plugin works like a visualizer, except it sends data out the parallel port as it interprets the music. On the website, there are a couple of different version of DiscoLitez. The Winamp plugin is called "DiscoLitez Standard". There is a Pro version that looks pretty cool, but I couldn't get it to work and didn't have time to debug.
I ran into a small problem with DiscoLitez at first. Every time I tried to use it, Winamp would crassh. It turns out that Windows XP has some security policy where software cannot talk to the parallel ports unless it has special permissions. I also had to use a program called PortTalk. PortTalk works by somehow giving the needed permissions to program that wouldn't otherwise have them. In my case, I had to give these permissions to Winamp. So, I launched Winamp through PortTalk. I made a batch file that contained the command line:
allowio "C:\Program Files\Winamp\winamp.exe" /a
This is a good guide for setting us DiscoLitez and Winamp.
I went with the Winamp/DiscoLitez setup mainly due to time constraints. It does the job, but does leave a bit to be desired. Here are some programs I saw while doing research. I may upgrade to one of these in the future.
Vixen - Very mature freeware program with many features and plugins. From what I saw, it uses the scripted model. Large community of enthusiastic users. I suggest you check out the wiki.
VSA - A $60 commercial application seen in other singing pumpkin rigs. They offer a demo version on their website for download. I did not test it myself. It appears that the software uses the interpreted model, and then lets you customize the script ... so I'd call it a hybrid model.
DiscoLitez Pro - Freeware. This is a standalone version of DiscoLitez. It appears to be completely interpreted model, but you have a lot more control over how the music is interpreted than with the Standard version.
Lightning - Freeware. Another interpreted model program very similar to DiscoLitez Pro.
Here are some future enhancements that I'd like to implement one day.
Figure out how to get the relay box to work with a USB parallel port. Newer laptops don't usually have parallel ports.
Improve illusion of singing.