This made a lot of sense. Audio books are a booming business these days, and it just makes sense to have your book available in as many formats as possible. Also, I’ve been an audio guy since the age of sixteen when I got my first job announce-operating at CJRW in Summerside, PEI, later making my living as an audio technician/recording engineer for CBC Radio for nineteen years.
For an entire ten of those nineteen years at CBC Radio I made radio plays and recorded and edited tons of short fiction re-purposed for the medium of radio. I remember recording a radio-friendly version of Brad Smith’s novel All Hat over the course of a week or two.
So you would think that I would know what is involved in such a recording. Unfortunately, all my experience did was give me a wildly over-inflated sense of my own abilities. Yes, I did (more or less) possess all the skills required to produce an audio book. But somehow I completely failed to appreciate just how much work was involved in doing it all myself, and how demanding some of that work was.
When Lorina suggested I do the audio book, I truly thought I would be able to knock it off in a couple of weeks. Because I could read, I could record, and I could edit. Thinking back, I was pretty sure we’d done All Hat in a week or two.
It’s laughable, really.
Because thinking back on it a little more carefully, I’m pretty sure that the version of All Hat we produced was an abridged version, and it took four of us to do it: a recording engineer, a producer, an actor, and somebody to adapt it. Five people, if you include the casting director. And all I did was record it (I may have edited it, but I don’t really remember). I certainly didn’t read it.
Anyway, turning my novel into an audio book was a great excuse to gear up, so I went out and bought a mic, a mixer, and some other peripherals. I had a week of vacation time coming up and figured I could squeeze all the recording in then, and edit at my leisure afterwards, on evenings and weekends.
After one week of recording though, I only managed to record ten chapters. My wife attributed this to my propensity to get up late, linger over breakfast reading the Toronto Star, casually walk the dog, and then get started recording around 11am. All of this was true. Add to that trains going by, planes flying overhead, neighbours noisily draining pools, and mysterious noises with no obvious provenance interfering with the recording when I finally did get around to it, and you can see why the process took a bit longer than expected. Worst of all, though, was my inability to read more than half a sentence without making a mistake.
Turning a novel into an audio book was a much bigger deal than I’d realized.
In fact, what I originally thought would take me two or three weeks to accomplish wound up taking over two hundred hours spread out over eleven months.
Here are a few thoughts on the process while it’s reasonably fresh in my mind, in case anybody else out there is thinking of doing the same thing.
To record my audio book, I settled on a Shure SM7B microphone. I chose this microphone because I had chosen it back in 2007 to be the main microphone for the radio show Q. I’d tested a lot of microphones and it had sounded the best with the host of that show, and it sounded pretty good on me (if I do say so myself). I would have preferred a Neumann U-87 but I couldn’t afford that (it’s about three grand). But the SM7B (at about $500 Canadian) is a fine microphone with an excellent pedigree. Michael Jackson famously used it to record his album Thriller. Its only limitation that I could see is that it’s a dynamic microphone and you need to give it a boost to get decent levels. But this is easily fixed by placing a Cloudlifter in the chain, providing an extra 25dB of gain.
An advantage of the SM7B is that it pretty much records what you point it at and rejects most everything else. This was really helpful recording in my basement. When I turned off the air conditioning, made sure no other appliances in the house were running, and closed the door to the basement, the noise floor was almost non-existent, but there could still be some extraneous noise, so it was helpful to have a very directional microphone.
You do have to work the SM7B pretty closely to get a nice, plummy sound. The host I used to work with on Q worked it so closely that I wound up sticking two pop filters between him and the mic to avoid popping. In my case, I used the A7WS windscreen that comes with the mic out of the box plus one pop filter. I still popped a bit, but I had ways of dealing with that, which I’ll come to later.
The rest of my setup was pretty simple. You can see it pictured here. Basically the SM7B plugged into the Cloudlifter, the Cloudlifter plugged into a Steinberg mixer, which in turn is connected to a MacBook Pro via USB. And a pair of decent Sennheiser headphones and a mic stand. I read the script (just a PDF version of the novel) right off the MacBook, flipping back and forth between Adobe Reader and my audio software as required.
I recorded almost everything in Cubase, which came with the Steinberg mixer, but I never really got to like it. I’ve used a lot of audio editing software in my time (D-Cart, Dalet, DaletPlus, Sonic Solutions, ProTools, Audacity) and Cubase just didn’t compare in terms of immediate usability. Probably if I’d taken the time I would have gotten used to it, but when it came time to editing the audio book, I switched to Audacity, which can be downloaded free and is much simpler.
Earlier I mentioned that I couldn’t seem to record half a sentence without making a mistake. This was true in the beginning, and it surprised me. One of the reasons that I thought recording an audio book wouldn’t take too long was because I figured I’d just sit down and read it and do some light editing and that would be it. I’ve had some experience acting and I’ve worked professionally as an announcer/operator at two radio stations (CJRW in Summerside and CFCY/Q-93 in Charlottetown). I thought I could read. Heck, I even thought I could perform. But I couldn’t. Not in the beginning.
The problem was I would read a little bit and then, convinced it sounded horrible, I’d stop and start again. I thought, well, not a big deal, I can edit it all later. But the more mistakes you make, the more editing is required, and eventually all that extra editing adds up to one big editing nightmare.
I got much better with practice and experience, but even at my best I couldn’t get through a chapter without a fair amount of mistakes.
Typically, I recorded each chapter twice. I would get to know the chapter on the first read, and read it better the second time around. If I made a mistake, I’d stop, go back, and correct it right away. This made the editing process much easier later (making up somewhat for the amount of mistakes).
Because I didn’t have a producer, someone standing over my shoulder correcting me, I needed to be careful. If I thought I made a mistake during a passage, I always stopped and re-did it (sometimes the first time was perfectly fine, but better safe than sorry, although it did make for more work). Whenever I hit a word I wasn’t entirely sure how to pronounce, I looked it up online. Most online dictionaries allow you to listen to the word you’re looking up. Interestingly, I included words in A Time and a Place that, although I know perfectly well what they mean, I either didn’t know how to pronounce, or have been pronouncing incorrectly. They are correct in the audio book version, though. I made sure of that.
Sometimes I mangled words or sentences but didn’t discover this until the editing process, which was a pain in the butt, but far from insurmountable. One of the advantages of recording an audio book yourself is that afterwards the actor’s still hanging around if you need him or her.
One of the fun parts of recording this novel myself has been doing the voices. I didn’t have unique voices for all of the characters, but some characters cried out for special treatment. One of the characters, Gordon Rainer, is supposed to speak with a British accent. He was by far the most difficult to get right. I’d once done a play for which I’d been trained to speak with a British accent, but I have no illusions about how accurate I’m able to do it (my British brother-in-law is only too happy to provide reality checks on that point).
I’d always thought of another character, Doctor Humphrey, as having a gruff voice, so I played him that way. And so on. I tried not to overdo it, as it could easily get silly, but I enjoyed the performance aspect of it all.
Like just about every other part of this project, the editing took a lot longer than I expected.
I edited one chapter at a time. It took me on average three to four hours to edit the first pass of each chapter. My chapters average twenty pages. The shortest is seven pages, the longest thirty. Transformed into audio, my chapters run anywhere from eleven minutes to thirty minutes long, averaging about twenty minutes. (Unedited, the raw files for each chapter run anywhere between one hour to two hours long.)
As mentioned earlier, I did all my editing in the free version of Audacity, which worked just fine. It’s easy to learn and I found that I could edit quite quickly and effectively with it. It’s also got a nice little suite of tools for mastering, EQ and so on.
Before editing each chapter, I would do a little processing. A little noise reduction, a little limiting or amplification as required to ensure that I was peaking at -3dB with a maximum -60dB noise floor as required by Audible. I did this at the beginning because if doing any of that introduced any problems, I wanted to catch those problems as I was doing the edit. I didn’t want to complete an edit, then do processing, and have it accidentally introduce issues such as clicks or pops or digital distortion that I might have less chance of catching near the end of the process.
Every chapter required multiple passes to edit. The first edit was mainly to get all the right takes in the right order and clean it up as much as I could. To help speed things up, I created a special template in Audacity’s EQ plug-in to eliminate popped Ps as I encountered them (I found the default EQ template for this too aggressive in Audacity).
Sometimes I encountered mangled words or sentences for which no good takes existed, that I had not noticed during the recording process. These I re-recorded right on the spot. Sometimes it was a bit tricky getting these re-takes to match, but it got easier with practice. It was a matter of getting the inflection and level right. I would tweak the level in Audacity using the Amplify plug-in (always careful not to peak at higher than -3dB), and try to use as little of the re-take as necessary, often cutting halfway through a sentence, or a word, even.
After the initial edit, I would go through the chapter again to clean up weird, extraneous noises such as bits of mouth noise, the cat knocking into the mic stand, or other weird noises such as bumps occurring elsewhere in the house that I hadn’t noticed during recording.
I took a lot of time to address issues with pacing. I tend to read fast. Left unedited, few would be able to keep up with my reading. I worked hard to address this in the edit. I know that some audio book listeners want their audio books read fast. In fact, they will listen to their audio books at enhanced speeds to get through them quicker. I tailored my pacing for people who listen at normal speeds. If I ever record another one, I’ll try to get it right during the recording. A lot easier than having to fix pacing in the editing process.
Once reasonably certain that I’d addressed all issues in the edit, I would play the entire chapter from beginning to end to make sure that I hadn’t missed any edits, and to ensure that there were no other problems. Only when I felt that the edit was perfect would I consider it done, and share it with my publisher via Drop Box (who will subsequently submit it to Audible). One of my mottos is “If it only exists once in the digital domain, it may as well not exist at all,” so I always sent a safety version to myself via Gmail.
I didn’t keep a really accurate record of how long this project took me, but I estimate each chapter took on average two hours to record, and six hours to edit, master, and double check. That’s 8 hours a chapter times 27 chapters, plus little bits like intros, acknowledgements, and so on. I figure the entire project took about 220 hours. That’s 27 eight hour days. The book itself is ten hours, sixteen minutes and fifty-five seconds long, all told. I took three entire weeks off work and devoted several evenings and weekends to this project. Probably much longer than it should have taken. I read in this quite informative blog post that “your narrator will put in six times more production hours than the final length of the book.” Yeah… took me a bit longer than that.
But I think it was worth every second.