Home -> Stephen Uitti -> Audio Books





Audio Books

Maintained by suitti@uitti.net, Stephen Uitti
How to create mp3 audio books for free (cheap).

Books over about 75 years old fall out of copyright in the United States. They can be copied for free. Project Gutenberg is an attempt to make such works available for free download on the internet. Originally, these works have been available in English in text form. Lately, the Project has expanded to other languages, and other formats. Some of these formats include HTML (styled text with pictures for a web browser) and mp3 Audio - with books read by a human or a computer.

There is also a site devoted to books read by humans.


In my opinion (IMO) it should be called AudioBooks4Cheap, since they charge a small fee for downloading reasonable quality audio. I mention them because they made several books available to Project Gutenberg, at least for awhile.

Audio Books For Absolutely Free

Find a book in text form, and download it.


Books are available in plain text form. They claim they have over 10,000 books on-line. One could just read them as-is. I'm fond of reading them on a PDA (such as a Palm Pilot). One could easily do this on a computer screen. To read them on a Palm Pilot, there are readers which require a special compressed format, and conversion software (and reading software) are freely available. I have run such software under Linux.

Next, one needs software to convert the text into audio speech. For Linux, I use festival:


There is no binary distribution, as yet. The source distribution is mirrored on a US site:


It appears that you need everything. You then extract all the files, compile speechtools, then compile festival.

The festival program converts text input to .wav format audio. mp3 files are much smaller, and my car's CD player can play them. Unfortunately, there is some very strange patent laws covering the mp3 format, and as a result, mp3 conversion software is not available in the US in binary form. You must either compile the programs from source, or you must download working binaries from someplace that allows it, such as Australia. The conversion utility that seems to work the best for this work is notlame. (gogo uses much less CPU, but the resulting mp3's generated are much larger. This is probably because it doesn't understand mono. The audio quality seems about the same. It runs on x86 architecture CPUs only.).


My car's mp3 player does not remember where you were within a track if you turn it off (for example, when you turn off the car). Also, these books are at least four hours in length. Also, festival has trouble converting whole books, running out of memory. So, the books must be divided into chapters. I use a text editor. Often chapters start with Chapter... and one can search easily. Once you have a test of files, a script, such as the following will run festival, then notlame on each.

# Convert a book (set of text files) to mp3.
# Current directory, *.txt => *.mp3
for i in *.txt; do
 echo $i
 A=`basename $i .txt`
 sed s/_//g < $i > $A.t
 time text2wave $A.t -o $A.wav
 rm -f $A.t
 time notlame --silent -h -mm $A.wav $A.mp3
 rm -f $A.wav

Then, I cut a CD. Most Linux distributions have mkisofs and cdrecord loaded. Red Hat 9 Linux has X-CD-Roast - a CD burning program with a nice GUI interface. I still use mkisofs and cdrecord.

My car player can read mp3 formatted CDs if the format is ISO 9660 with Joliet extensions. This is what Windows would write for a data disk.

One should note that all DVD players today can read mp3 CD disks. I have mine hooked up to my stereo, and I can play audio disks (with or) without the TV on.

I'm finding that I'm getting 6 books in audio format on a 700 MB CD. That's 70 to 80 hours of Audio. For music, I usually get about 14 hours. Spoken audio is one track - mono, where most music is stereo. That alone doubles the time on a CD. Spoken audio is easier to compress than music. There are lots of blank pauses between words that compress to essentially nothing. I also get this kind of compression performance with very low quality human spoken audio.

You probably won't want to split a book across two CDs. With books requiring between about 60 MB and 150 MB, there may be considerable unused space on a disk. One can put the text versions on the disk without difficulty, since the text for the six books will occupy about 2 MB, uncompressed. CD players ignore files in formats that they don't understand.