Text files are often limited to a small amount of characters. Often just linebreaks, spaces, the alphabet in upper and lower case, numbers and some other special characters. In general, a textfile will use about 100 of the 256 possible values that a character can have. Basically, since a character is 8 bits and only the lower 7 bits are used in textfiles, the compression algorithm can already remove 1/8th of the file size by storing the data in 7-bit format. That already reduces it by 12.5%.

Media and audio files tend to be compressed by default already. In uncompressed format, a single minute of spoken text could easily be several megabytes in size, depending on the quality. (*.WAV files tend to be stored uncompressed this way.) However, many of the file formats used will do some compression on the data to reduce the file size and they tend to have pretty big compression rates too. A WAV file of 50 MB could be reduced to a tenth of it's original size or even less by storing it as an *.MP3 or *.WMA file.

Compare compression with folding a piece of paper. Normally, you'd need a large surface to store a piece of paper. By folding it several times, you reduce the size (but make it thicker) so it's surface gets smaller. But you will get at one point where you can't fold the paper anymore because it gets too thick. (Somewhere around folding it 9 times or so.) The same is true with data compression. A file can be 'folded' up to a certain size but it won't get any smaller than that.

See also this WiKi page.