On the importance of proof listening
Updated: Sep 21, 2021
The creation of an audiobook is an enjoyable, multi faceted process. Once you landed the gig, it's time to prepare the manuscript - not only so you know what's going on and who's involved, but can also brace yourself for fun sentence structures and words you may have never heard of before. Or words that are pronounced differently than you thought for your entire life up to this point.
So you do your due diligence during prep and give it your all during the recording. And I am sure, you had every intention to make no mistakes (or get rid of them with Punch & Roll right away)!
I hate to break it to you though - chances are, that some inserted, dropped, flipped word evaded you and made it into the finished recording.
So, proof listening is essentially the Quality Assurance for the content you are delivering - though not for whether your files meet the submission standards of any given audiobook distributor.
You'll find out where the recording doesn't match the text, if there's a word that wasn't pronounced quite right, where you might have any audio artifacts or background noise and if a line of dialogue got the wrong character voice.
So, there are a couple of ways of going about proof listening:
Skipping proof listening altogether. This is something I would recommend against doing. There are so many missed opportunities when you skip this step, which can really hurt how your final product is perceived by listeners and potential clients.
Do it all by yourself. When starting out with your first royalty share gigs, this is probably the approach you'll choose (for budget reasons, most likely), and I'm not going to lie - early on this is what I did, too. The biggest challenge with this is that you're too close to the project and too familiar with it. Chances are that you might miss especially small word flips or mispronunciations because they feel natural to you. There is that text floating around with jumbled letters indicating that you can still understand the text because the brain pieces the puzzle together automatically. That's kind of what's happening with flipped words. And pronunciation - well, if you think the word is pronounced a certain way, you won't catch it being mispronounced.
Use an automated service - for me it's Pozotron. First off - I am a fan of Pozotron. It's helped me out a lot already and now I feed pretty much all my texts through it. As you create a new project, you upload your script for its own reference. Then you upload the various recordings from the project - clean punch and roll recordings, so any retakes need to have been taken care of - and it analyses and scans the recordings for any issues. It matches the audio to the text, highlighting the word that's currently read. It will very very very likely catch words that have been flipped, dropped, inserted etc. So you may very well augment the DIY approach with pozotron and be fine for a bit, but it is not perfect - so don't expect it to do ALL the work for you. Aside from "false positives", there are things it just doesn't catch as well as human ears do (including some pronunciations - think wound/wound, bow/bow, read/read etc.), but it can help you in the DIY approach. If you wanna just try it out - this link lets you upload 7 hours of audio to play with for free, no strings attached.
Work with a human proof listener. First of all - the author or rights holder of the project you're working on should not be your proof listener for the same reason why you yourself are not the best proof listener - you're too close to the project :) Once you are able to budget for a human proof listener (that is not you) - or can talk your dear friend/family member/partner... into helping you with that part, this is a big step up. The distance from the project means that the PL listens to it with fresh ears while following along with the manuscript. They'll also be quicker to note if a pronunciation is off in general or in this specific context, will point out unwanted audio artifacts etc. and create a handy list (proofing report) for you to work down. And once you fix the issues, they'll do spotchecks to ensure you got it right. I'll admit that both "countenance" and "coquetry" took a couple of tries for me. With one I kept falling into a French-like pronunciation and with coquetry I kept switching between British and American English. Ah well. Oh, right! The human proof listener is also part of your team. I work with Listen Closely Proofing a lot and, at this point, consider her a friend with whom I can talk about things beyond any given project :) Audiobook narration can be a pretty lonesome profession, so having someone on your team is worth a lot. The downside is that sometimes a mistake might even slip by the most attentive human PL's ears, too. It's the whole thing about our brains jumbling jumbles together for us and trying to be helpful.
Work with a human proof listener AND a software based technology. This is combining 3 and 4 and turning the whole proof listening thing into a power house. Aside from the other features that come with, in this case, Pozotron, the technology to highlight the currently spoken words alone help follow the text along - something that also makes sneaky misreads visible (if, for example, you're saying 'she says' and the text asks for 'says she'). The PL can then add annotations of their own, dismiss false positives from Pozotron or accept things the software found. You essentially combine the power of human ears (pronunciations, character voices, audio artifacts etc.) and the power of totally neutral software analysis, giving you the most accurate picture of which pickups you need to take care of.
And you know what? Having to do pickups actually gives me the opportunity to notice phrases that I didn't love the read of, too. Because usually I have the most pickups when my focus starts waning a little, and later I realize that those phrases could be read better, differently. I don't do it for every pickup, but to take the opportunity when I can.
As for the pickups - personally, I like to listen back to the area in question, and then record a little bit before and after anyway. That makes it a little easier to match the expression of the original, which helps a seamless patching in of the pickup. Or if it just won't fit - I can replace a little bit more. Then I make sure to note the timestamps where the fixes can be found and tell my proof listener about them - chances are that my fixes changed the original timing a bit (I may have accidentally left out an entire line in the past. Don't tell anyone. ;) It did get caught and fixed thanks to this proofing stage.)
One easy way to do it, if your recording software / DAW allows it is to just set markers for where you patched them in, and then simply copy the times over from the marker overview. For example, in Adobe Audition it looks like this - note the timestamps in the middle.
So! Happy recording! :)