Tysto home


f r o n t . p a g e


b u s i n e s s


c u l t u r e


e n t e r t a i n m e n t


g o v e r n m e n t

e - m a i l . t y s t o


a b o u t . t y s t o

s e a r c h . t y s t o


Recording audio books

2008.01.06 — Entertainment | Books | Audio | by Derek Jensen

Audio recording setup

Tysto International Studios. [Tysto photo]

I love listening to audio books. I've been a customer at Audible for years and have also found some great classic (that is: public domain) works to enjoy at LibriVox, where the books are free and contributed by users. On the forums, I've seen posts by long-haul truckers, students, housewives, and others who listen to the free audio books and want to participate. Some readers are very talented, and so, being creative, bored, and competitive myself, I decided to try my hand—or mouth. That is, to put my mouth where my... mouth is. Whatever.

Take a look at my contributions to LibriVox.

The LibriVox way

The great thing about contributing at LibriVox is that there are several ways to do it, and the amount of time you spend can be small or large, depending on your availability. You can be:

  • A reader (who records and edits audio files and uploads them for inclusion)
  • An editor (who helps less technical readers edit and polish their recordings)
  • A proof listener (who looks for mistakes in the reading or editing)
  • A book coordinator (who organizes the participation of readers to complete books or collections of short works)
  • A meta coordinator (who catalogs completed works)
  • Or any combination thereof


Work is coordinated thru the LibriVox forums. Typically, the book coordinator will set up a forum thread for a book and designate sections of a chapter or two each and solicit readers. A reader will volunteer for a couple of chapters, record and edit the reading, and upload the file. The book coordinator will find listeners to proof listen and approve. Then the meta coordinator will shuffle the files and perform the final steps to catalog them as a finished work. The result is a long work with a dozen or so different voices reading different pieces.

In short work collections, readers volunteer poems or short stories of their own choosing, generally with a theme, such as ghost stories or science fiction. In special dramatic works, the coordinators will actually parcel out individual speaking parts to different readers, so the voice for Hamlet, for example, is consistent thruout the work.

In all cases, the source material must be in the public domain (under US law: published prior to 1923 or fallen out of copyright for other reasons). Moreover, the final product is also released to the public domain, meaning that anyone can use them, sell them, chop them up, or whatever, any way they want. This differs markedly from Project Gutenberg, where every file is heavily laden with legalese so that you can't sue them if you somehow choke to death on one of their e-books or get a B for misquoting a passage with a typographical error.

And typos are very common in these e-books. Most of them are scanned and transfered electronically, so "worn" can easily become "wom" and so on. I once ran into the phrase "he put his hand on his month" and puzzled over it for several minutes before deciding "month" should be "mouth" instead.

Tysto International Studios

Audio recording setup

Tysto International Studios, setup 2. [Tysto photo]

My audio setup is pretty simple and not a permanent installation. I have a Samson CO3U microphone ($90, plus $18 for a floor stand) for high-quality recording directly to a laptop thru a USB port (podcasters and others formerly had to use a professional-style mike and go thru a small control board).

For my first contributions, I was using a Logitech USB desk mike ($22), but it captured too much ambient noise and didn't have very good bass response. At Tysto, we do film commentaries now and then, so the extra money was worth it.

I run the wonderful free audio software Audacity to do the recording, editing, and output to MP3. And I listen with high-quality Sony earbuds ($50) to ensure I can hear all the little lip smacks and mouse clicks that I want to edit out. (Tip: The earbuds are also useful for listening to audiobooks!)

Microphone level

It annoys me that, even with a wireless mouse several feet away from the microphone, I still capture every click, so I have to be careful about mousing around while I'm reading. Doesn't anybody make a silent mouse? What do podcasters use?

I bought a big, widescreen, secondary monitor a few months ago that I find enormously useful for photo editing, and I've begun using it to display my reading material while audacity runs on the laptop's screen. That allows me to set the laptop out of the way, reducing fan noise in the recording and still go back and forth quickly.

I highly recommend the Dell LCD monitors that rotate to portrait and landscape modes. I have the 20" widescreen, which has a resolution of 1680 x 1050, which means that in portrait mode (1050 x 1680) it can display a webpage designed for a 1024 x 768 monitor but show about 2.5 "screenfuls" without scrolling. Meanwhile, in landscape mode, it has the width and detail to make spreadsheets and DVDs—and landscapes—look great.

I do my recording in my basement home theater, which is pretty quiet, altho not always quiet enough. When the furnace is running, it's noticeable in the recording even thru the insulated walls, but fortunately, Audacity has a good noise-reduction feature and forced-air noise is easy to remove. I've considered recording on the second floor, but it wouldn't be as comfortable or convenient, especially since the Tysto film commentaries are naturally done in the theater.

Microphone level

I copy the material and paste it into MS Word, then format it for convenience and add the LibriVox header and footer, so I have a kind of script to work from. Sometimes, I even note which character is speaking by changing the color of the text for each character; or I bold certain phrases that I need to emphasize to make a tricky passage easier to read. I enlarge the text to about 200% so that it's easy reading, then scroll as I read, so it's almost like a teleprompter.

I find that I spend 8-10 hours producing 1 hour of audio book reading. That seems excessive and is actually more than the time it takes to record and process a 2-and-a-half-hour film commentary—and I'm thinking up all that material myself. Maybe I'll get faster as I do more books, but I've already produced several hours of audio for LibriVox.

Basic sound production

Microphone level

How to set your audio level input in Windows XP. I keep mine pretty low to keep ambient noise low and not overpower the mike with my voice.

I use Audacity, which is available for free for Windows, Mac, and Linux, but Mac users have similar features in Apple's Garage Band, which ships with Macs. I record at 44100 hz and export the final result to MP3 format at 128 kbps, since that's the LibriVox standard. This eventually gets converted to 64 kbps automatically and offered both ways, so that casual listeners can have smaller files (64 kbps) and those who want high-quality files (128 kbps) to use for other purposes can have those.

I do a couple of test recordings to ensure that my mike input level and microphone placement are good, then I record the file, generally in one or two tracks, pausing when necessary when the reading is long. Most regular readings are around 40 minutes; the maximum recommended is 60, since LibriVox wants to ensure that listeners can put files on CD for convenience. With all the MP3 readers available, I doubt many people use audio CDs, but it's a good rule of thumb.

It's good to keep your head up during the reading to keep your throat clear and good to have a beverage handy but not food.

It's important to re-read a passage any time you stumble (just re-read and push on; don't stop the recording to edit right then) so that later you can edit out the mistake or poor choice of inflection and get consistent sound in the next bit. Good advice I've found is to back up to the last piece of punctuation, rather than just re-reading a word or phrase, since words and phrases are often slurred together even in the most precise speech.

Basic sound editing

Once I've finished the reading, I go back and listen to it, reading along again to find those spots where I stumbled and re-read a passage and the find the spots where I read a phrase wrong and have to re-record. It's sometimes very hard to get the sound consistent again when you try to re-record something an hour later or a day later. The mike is in a slightly different position, your body is in a slightly different position, your vocal tone is slightly different, the ambient sound of the room is slightly different... it can be maddening. This is no big deal if it's a whole paragraph, but it's very, very tricky if it's just a few words.

From time to time, you'll find there are places when you'll want the start of one read and the end of another. Finding the right spot to join them together takes experience and trial-and-error. While joining at a silence is always best, I find that joining in the middle of an F or S sound usually results in a perfectly natural result, but joining in the middle of a vowel sound almost never does.

Audacity Change Tempo

For overlong breath pauses, select the pause and select Change Tempo from the Effects menu and increase the tempo by 50% or so.

One area of difficulty is breath pauses. Breath sounds (often accompanied by lip-smacking), tend to be mildly distracting, so some readers use the amplify feature (with a negative number of around -10) to reduce the sound at that point. My pauses tend to be overlong, since I take a second to double-check what I just read and to read ahead a phrase or two. My old technique was to delete any silence and keep the breath, since deleting part of the breath makes for an awkward sound. My new technique is to select the whole pause and use the change tempo feature to speed up that section (tempo changes the speed without changing the pitch, so you don't sound like a chipmunk). I set it for a 50% or 60% speed-up and sometimes reduce the volume while I'm at it.

For very long pauses (swallowing, resting, reading a tricky passage to myself to make sure I have the sense), I delete the middle and leave the start and end, so that the end or beginnings of words or breaths aren't cut off (S sounds and W sounds are often quiet enough that they look like silence in the display).

Audacity Leveller

To even out the volume a little, select the whole file and select Leveller from the Effects menu and specify the amount. Use too much and you'll introduce a harshness in the louder parts.

I finish off by doing some overall processing. First, I use the leveller feature at the light setting to raise the volume on the softer parts relative to the loudest parts. Then I use the normalize feature to automatically raise the peak volume in the file to the full volume level (so that the loudest part of the recording fills the boundaries of the track). Then I usually perform some modest noise removal. I select the quiet section at the end (which is supposed to be silent) as my noise profile, then apply noise reduction of 15 dB. Greater noise removal tends to result in weird "tinkly" artifacts, so use it sparingly. Similiarly, using the leveller at more than the moderate setting can result in distortion in the louder sections.


When you get good sound and a clean read and have edited out the slips, the result can be very satisfying. I don't have a particularly good voice, but I think my reading is very expressive and not hard to listen to. There are some LibriVox readers who are, or could be, professional voice actors, and the difference between them and the rank amateurs if often striking, but mostly in the audio quality rather than in the quality of reading.

I always use the LibriVox uploader (a link is available in the LibriVox "How to Record" instructions). It returns a link that I can paste into the forum thread so the book coordinator and listeners can find it. If there are any suggestions for change (missed a sentence, duped a phrase), I can reupload it easily and the link remains valid. Once all the parts of the project have been recorded and proof-listened, the project is closed and the work is posted in the LibriVox catalog.

It's a slick system LibriVox has put in place, and the world is richer for it. Unlike Project Gutenberg and their team of lawyers, they ask for no donations (they've managed to get free server space from Web providers). I don't think LibriVox is going to ever reach its stated goal of making all public domain books available on the Web for free, but the hundreds of volunteers and dozens of regulars are setting a heck of a pace trying.


f e e d b a c k

Respond to this page by your e-mail client. Please be sure to mention the title of the article.


s i d e b a r