IV. The Technical Aspects

Extracting the Audio

The first task in setting up any radio station is generating a playlist of songs. For games, that means extracting the audio into the common currency of MP3. Unfortunately, every game approaches audio differently. There are three broad classes of music that all games will fall into: MIDI, digital mixing, and CD Audio.

  1. MIDI
    Generally only found in older games. MIDI files are like sheet music; they only specify what notes and instruments are to be played back. The difficulty with MIDI is twofold: first, the quality of the audio will depend on the playback device, and second, you must capture the playback into a digital file in order to create a MP3 file out of it. After months of searching fruitlessly, I was fortunate enough to stumble across a fantastic utility that addresses both issues: timidity++. There's a nice walkthrough on how to use it at this site. It's still an involved process, so consider yourselves warned, but at least it's possible.

  2. CD Audio (redbook)
    Your basic CD-audio soundtrack on tracks 2 and higher of the CD. By far the easiest of the three to extract. The first thing I do with any game CD is slap it in the old CD player and see if it plays. If it does, any number of free or commercial products are available, which can easily "rip" the CD-Audio tracks to MP3 files. My personal favorite is the now-defunct AudioCatalyst, but I also hear good things about MusicMatch Jukebox.

  3. Digital Mixing
    This is sort of a catch-all term for anything that isn't MIDI, and isn't CD-Audio. In this case, the developers decided to take raw audio files and mix them via some proprietary method. In some ways this is the best of both words; it offers the quality of CD-audio, with the possibilities of dynamic arrangements of MIDI. But it's bad news for extraction. There's no good rule of thumb here on how to proceed; it will vary on a per-game basis. There are a few freeware tools that you can use to try a brute force extraction of the audio files, such as GAP. That's assuming the developers are using semi-standard file formats. Like I said, no guarantees here. If the game is popular, it's likely that some hacker has already come up with a way to extract the audio. For example, there are many extractors for Unreal's proprietary audio format, UMX.

Setting up the Server

Attempting to set up a radio station on a web server can be an unnecessarily complex endeavor. There's certainly no shortage of people trying to convince you to use their proprietary streaming software, such as Real's RealSystem, Microsoft's Media Server, and WinAmp's Shoutcast. Some of it is free; some of it costs thousands of dollars, and it's all complicated.

My philosophy is, when in doubt, keep it simple, stupid. With that in mind, I decided to forgo all the complicated streaming solutions and go with the most basic method: serving individual files via HTTP. We serve playlists for WinAMP and Windows Media Player. The playlists are simply text files, formatted as specified for each player, containing hundreds of standard URL links to songs.

Since this is a web-based project, cross-platform portability is a concern. Fortunately, MP3 files are universally supported on most platforms, as are the URLs used to serve the files. And WinAMP, at least, has clones such as MacAmp and XMMS (for Linux) that can parse the playlist files. So we're set on that front.


Windows Media Player
WinAmp

The other major issue was bandwidth. All the music tracks are standard 128kbps (or higher) MP3 files. These are great for listening to on your local PC, but not so hot for sending across the internet. There are two considerations here.

  1. Streaming requirements. All modern media players begin playing the audio files before they are completely downloaded. For this to work, the encoding rate cannot exceed the bandwidth of the person listening to the file. In other words, you can't expect someone on a 56.6 modem downloading at 50kbps to be able to stream a file encoded at 128kbps. A bit like pounding a square peg into a round hole, really. And when that happens, you have to waiting for an entire music track to download before being able to hear anything. This is tedious, as you can probably imagine.

  2. Bandwidth costs. 128kbps MP3 files are huge— an average audio track is anywhere from four to five megabytes in size. That is a hefty download, especially if you have 10 or more listeners at any given time. We aren't running a porn site here; even if we were living in some kind of ideal world where every listener has a cable modem or T1 line, I doubt I could afford to pay for the bandwidth we would be using in that case.

The solution to both problems is simple: resample the source files at a lower bitrate. You can resample MP3 files using the freeware tool, LAME (download binaries here ). I did a lot of experimentation and found the ideal bitrate to be 40kbps . Unfortunately, 40kbps MP3 files sound pretty crappy— roughly AM radio quality. Out of frustration, I experimented with alternate formats that both Windows Media Player and WinAMP support, in particular WMA (Windows Media Audio). In side by side listening tests, I found that WMA offers dramatically improved sound quality at the low bitrate I had decided on. Quite a surprise! After researching this topic further, I found many other outside sources that confirmed my results:

Though WMA offers a definite advantage in quality, it suffers from the predictable disadvantage of being completely Windows-specific. In order to maintain cross-platform compatibility, I would have to maintain two sets of files: one with all the songs downsampled to 40kbps MP3 format via LAME , and one with all the songs downsampled to 40kbps WMA format via AudioConverter . This essentially doubles the amount of work for each track, but it's necessary if we want to serve listeners on non-Windows platforms.

 

 

Back to Features