Continued from last week…
We’ve received a lot of questions lately about our recently released AI-narrated audio books, mostly revolving around why we did them and what the process entailed. Before we go any further, here’s the list of books we’re talking about. Free samples are available at the links. I’d encourage you to go listen to at least two samples—say, Emerald of Earth and The Counterfeit Captain—before forming an opinion.
Amazon AI-generated “Virtual Voice” narration
Emerald of Earth
The Midnight Ground
Hart for Adventure
The Recognition Run
The Recognition Rejection
The Recognition Revelation
Actual Living Human narration
The Counterfeit Captain
The Fugitive Heir
The Fugitive Pair
Our reason for doing the AI-generated audio books was simple: we were offered the chance to participate in an Amazon Beta program, and we thought we’d give it a try. We’ve been interested in adding more audio books to our lineup for a long time, but The Counterfeit Captain, The Fugitive Heir, and The Fugitive Pair were hellishly expensive budget-busters to produce, and the voice actor we booked to do The Midnight Ground never did deliver finished and usable files.
So we figured: why not give this Beta a try? We can always pull the audio books from release if we don’t like the results. We picked Emerald of Earth as the first test subject because the idea of using Amazon’s AI to narrate a novel about an evil AI bent on world domination was just too ironic to resist.
When Amazon’s AI generated and published the 8-hour-long audio book in less than two hours, as compared to the weeks it took to produce each of our human-narrated audio books, that was when we decided to generate audio books for The Midnight Ground, Hart for Adventure, and the Recognition Run trilogy, to see how they turned out.
Q: Will you be producing audio books for Stupefying Stories?
A: No. We don’t buy the audio book rights to the short stories we publish. Therefore, AI narration or not, we won’t be publishing audio book editions of Stupefying Stories. We do have an audio book edition of The Odin Chronicles in the works, but this was created using living human narrators.
I would like to put together a collection of short stories narrated by their authors, but the logistics of doing so make this a blue-sky dream, not a practical project.
Q: How did you select which books to convert?
A: Amazon did the initial selection based on something intrinsic to the epub files to which we are not privy. They presented us with a list of our books they said were suitable for conversion. We selected the ones we wanted to convert.
Q: What can you tell us about this Beta program?
A: Not much, I’m afraid. I’m sworn to secrecy. I can tell you though that this does seem to be a real Beta program. They’re frequently asking for feedback, and do seem to be taking our input seriously. I’ve already seen two of the issues I’ve raised addressed in program updates. Every time they issue a new update, they invite us to re-generate the books, to take advantage of the new changes.
Q: How much control do you have over the virtual voices?
A: At this time, not as much as I’d like. Amazon initially provided us with eight virtual voices from which to choose, which amounted to these polarities: male-sounding or female-sounding; American accent or British accent; and “tense” or more relaxed sounding. (Yes, that makes eight choices. Two to the third power, you know: 2 x 2 x 2 = 8.) We were required to select one narrative voice for the entire book, and had some limited ability to fine-tune pronunciation on a word-by-word basis, but not much.
The first thing I asked for was the ability to assign narrative voices by chapter, rather than by book, because we have some books in which the story is told from alternating characters’ viewpoints. The second thing I asked for was greater age range in the virtual voices: right now all the A.I. voices sound middle-aged. I’m told both these changes are in the works and will be rolling out later this year.
Q: Wouldn’t it be cool to have [insert your favorite wish-list item here]?
A: Yes, it probably would. But remember, this is a Beta program, and the objective here is to use the power of A.I. to generate a serviceable audio book straight from the Kindle e-book file without having to do a lot of extra markup coding.
I for one would dearly love to have a full and proper audio book markup language, incorporating elements of both IPA and music notation, so that I could really fine-tune the pronunciation, pitch, pace, timing, dynamics, and delivery of the lines by the virtual voices. But having that would require us to produce yet another source file, which would blow the whole point of being able to produce the Kindle, print, and audio book editions all from one common source file.
At the same time, I realize that having such a markup language would really play to my perfectionist tendencies. Remember, I’m the sort of person who could spend six hours in a recording studio trying to get the perfect three finished minutes on tape. If I had such an audio book markup language, I would waste ungodly amounts of time trying to get the narration absolutely perfect, and that would be a huge mistake.
I have written both scores and scripts. Part of the charm of working with living performers live and in real time is that you don’t have total control over what happens on that stage; you can only provide your performers with a framework for what you want them to do and then hope they follow it. Until A.I.s can flub a line or miss an entrance and then try to ad lib and improvise their way out of it, A.I.s will never truly be able to replace humans.
Q: How come there’s a photo of cows at the top of this column?
A: To illustrate a point. I took that snapshot with my phone the other morning, and to my surprise, as soon as I framed the shot, my phone recognized the scene as a pastoral landscape and launched an A.I. filter to “enhance” the photo by making the grass look more lush and green and smooth-out the colors of the cows. I didn’t even know my phone’s camera had A.I. software in it. It didn’t used to have it. It must have been something that was downloaded and installed in the last software update.
This is the way A.I. is going to come into your life. It’s not going to be a single thing that sits up one day and says, “Cogito, ergo sum.” It’s going to be a swarm of little enhancements and apps that slip into your life quietly through the back door, and then patiently wait for opportunities to start making subtle but significant changes to everything you see and read…
And write.
If you like the stories we’re publishing, become a supporter today. We do Stupefying Stories
out of pure love for genre fiction, but in publishing as
in tennis, love means nothing. To keep Stupefying Stories going at this level we
need to raise at least $500 USD monthly, and rather than doing so with
pledge breaks or crowd-funding campaigns, we’d rather have supporters. If just 100 people commit to giving $5 monthly, we can keep going at this level indefinitely. If we can raise more, we will pay our authors more.
0 comments:
Post a Comment