slightly more perverse, if you're interested in the space... I happen to know this guy from high school and the stories behind his performances are fascinating
I have a lot of respect for voice artists who read out whole novels especially. I recorded a blog post of mine a while back and it was surprisingly really difficult to get right.
A great narrator can create an audiobook that transcends the text. A terrible narrator can render an audiobook unlistenable. I suspect the robots will eventually take over (especially if you can tweak voices to your taste), but until then I'll continue to enjoy the likes of Steven Pacey.
Text-to-speech seems to have hit a certain plateau where it's perfectly functional (for English at any rate). But it would probably take a lot of incremental work to get to network news anchor much less great audiobook narrator territory. There's probably a great deal more value in improving speech-to-text at this point given that AI transcriptions are better than nothing but still quite poor relative to humans.
I don't doubt that it'll take a lot of incremental work, but look at the rate of progress everywhere else. In 5-10 years, AI speech - coupled with good text models that can understand sentiment and subtext - will probably blow us away.
I don't believe it, but I suppose anything's possible.
The narrator has to be a believable person when he/she reads the dialog. I guess that will be the first barrier to fall, followed by a visual AI "person."
It's already possible to tag (parts) of sentences on how tts has to speak them, quiet, excited, etc. Soon it'll be just a matter of what is cheaper, a recording or a tagger/director, and that tagging will be partially automatic. As with many things absolute quality, sadly, won't be the target, good enough to keep people buying the product is.
Not sure how I feel about this prospect, I love audiobooks, and for big titles the narration cost is a drop in the bucket, and I like the idea of real narration. But at the same time lots of indie books don't get an audio version, and that tts allows more books the have an audio version is a plus and having each character get their own voice works be nice.
I think there will always be a premium for human-sourced media. People are going to love being able to tweak the cast and narrator for their favourite audiobooks, but there will always be a market for your favourite celebrity's reading as well. People will swear they can tell the difference between a human versus a machine rendition, and debates will rage the same way they do with audio- and oenophiles.
Either way, we'll see within a decade, I expect. I would argue it's much, much simpler than self-driving cars - the text models of today already understand (or at least can classify) a great deal of subtext and sentiment, and the actual audio rendering has become extremely listenable on short texts.
I make my living as a voice actor. I've done everything from audiobooks, commercials, ADR to corporate.
The best thing you can do is practice, practice, practice, but with good guidance.
The first thing is to love and admit your voice. A lot of people try to sound deeper, more impostured, etc. They try to put on a voice they don't have. And the first thing is to recognise the sound of your voice. Your voice is your voice. You don't have to do anything special with it. Just accept it, understand how it works. It's that easy, it's that hard.
There are plenty of schools that teach voice acting. It is a difficult profession to be self-taught, as it requires an ability to "isolate yourself from yourself", as well as being directed by an outside observer. It is exactly the same as learning to act. You can learn from videos, books, but there is no substitute for physical learning with another person.
Audiobooks are one of the most difficult genres to work in.
They require you to tell the story, but you can't be more of a protagonist than the story. In most of them, I listen more to the narrator than to the story itself, which is a problem, as it is the story that has to transcend.
If you are curious, one of the classics to read are Cicely Berry's books "Voice and the actor" and "Freeing the natural voice" from Kristin Linklater.
If I may ask: How did your journey as a voice actor start, and how did you happen to make a living off of it?
I love to read aloud for decades, do a bit of different voices per character... And people seem to like my voice when I moderate our daily rounds or do presentations. I just learned that in the last 12 months, so maybe there is a hidden talent that needs to be unveiled.
I got to know the world of voice acting because I am a sound engineer and I had to record quite a few people. I started when I took acting classes, but the truth is that since I was little I liked to clown around with my voice. I've taken TONS of classes over the years.
There are so many genres where a voice actor can do voice acting. Think of anywhere you hear a voice. There is a voice actor behind it.
I don't know what area you live in but if it's something you're interested in, I'd look into where you can take classes.
I have never bought an Audible book without listening to the entire five minute sample. Narration is super important.
I mentioned to some friends that my wife and I have about 500 Audible books between us on our family plan, that is not counting audio books listened toon Libby from our local town library. I spend easily ten hours a week listening to non-tech audio books a week but only about five hours reading non-tech books.
Several years ago I read James Joyce’s Ulysses and to be honest I didn’t really “get it.” A year later I listened to the audiobook that was narrated by a half dozen skilled actors and the whole thing came alive for me.
Do audiobook narrators read the exact same book that the public reads, or are they annotated/sorted by characters/etc? More of a script? I've always wondered how they keep characters separate so well. Is it something as simple as different font colors per person, etc?
I've always been curious about the logistics behind getting an audiobook out.
I don't know what a big time publisher does, but in my case, it was just a pronunciation guide for Max, plus reviewing his first draft. Yes, I had to listen to every single second, at 1.5x speed.
If the narrator can't tell who is talking, the reader won't be able to either. I think there was only one case where he had to guess, and guessed wrong.
Narrating a technical book: some Shameless Self-Promotion here. Two books for that hard-to-shop-for person on your gift list.
I considered, briefly, narrating my own books, but then after a little research, realized that was a bad idea. It is acting.
I auditioned about 25 actors, both on ACX and on Voices, with a 2-minute section from my book that has all three main characters talking (two male, one female). I really wanted to hear how they did it.
(On this one, there is a section of Mesa code. I had to tell him exactly what to say, which I modeled on what a CS professor would say if he was reading off the whiteboard.)
What was interesting for me was, on Voices, some of the actors only gave me 20 seconds, and said they'd do the rest if I wanted! I found this incredible: it's only two minutes, for God's sake. I need to hear how you do the voices.
In their market, some producers will listen to that 20 seconds, say "Nope!" and turn them off. So they're reacting to that. Sad.
Maybe if you don't have people speaking, you can do your own book. Otherwise get a voice actor.
Edit: if anyone would like the text for the samples on the Amazon page, just speak up.
interesting, when I used to read more fiction, I never actually 'read the words' either, I just skimmed the page with my eyes and my mind played a movie in my head
wonder if the same pathways can work for audio, do you use something custom to read for you or something off-the-shelf?
Voice acting is a professional career. I worked with lots of voice
talent during my time in entertainments. Like all professional actors
they put a lot of effort into developing their range and characters.
I recorded the audio-book of Digital Vegan myself. I am a terrible
voice actor but have a good deal of experience with public speaking
and radio work. Author-narrated works, especially if technical, have a
couple of advantages in that specialist domain knowledge (acronym
expansion, pronunciation etc) comes through confidently, and there is
some evidence that audiences like author-reads for authenticity.
What I discovered when trying to distribute the work shocked me, and
that's why I ended up just putting it on my own site (money was never
really the main objective).
The market is dominated by five or six US entities that aggregate
smaller distributors and online stores. Buried in the final stanzas of
two of five contracts I reviewed were clauses signing over additional
rights to machine learning training for the audio recordings plus
e-book transcripts. One was actually on a page alone, breaking with
the typography and structure of the rest of the contract and following
a blank page, so clearly intended to be "snuck-in".
I won't name the companies for legal reasons. But voice actors beware.
Read every line of the contracts and if you work for hire for authors
it might be worth explicitly excluding permission for machine learning.
Not an issue. You (the author) give them a pronunciation guide, and then you listen to their first draft and make corrections. Possibly even record it yourself for them, as a help.
26 comments
[ 4.0 ms ] story [ 74.1 ms ] threadhttps://www.menshealth.com/sex-women/a35031293/joe-arden-nar...
The narrator has to be a believable person when he/she reads the dialog. I guess that will be the first barrier to fall, followed by a visual AI "person."
Not sure how I feel about this prospect, I love audiobooks, and for big titles the narration cost is a drop in the bucket, and I like the idea of real narration. But at the same time lots of indie books don't get an audio version, and that tts allows more books the have an audio version is a plus and having each character get their own voice works be nice.
Either way, we'll see within a decade, I expect. I would argue it's much, much simpler than self-driving cars - the text models of today already understand (or at least can classify) a great deal of subtext and sentiment, and the actual audio rendering has become extremely listenable on short texts.
The best thing you can do is practice, practice, practice, but with good guidance.
The first thing is to love and admit your voice. A lot of people try to sound deeper, more impostured, etc. They try to put on a voice they don't have. And the first thing is to recognise the sound of your voice. Your voice is your voice. You don't have to do anything special with it. Just accept it, understand how it works. It's that easy, it's that hard.
There are plenty of schools that teach voice acting. It is a difficult profession to be self-taught, as it requires an ability to "isolate yourself from yourself", as well as being directed by an outside observer. It is exactly the same as learning to act. You can learn from videos, books, but there is no substitute for physical learning with another person.
Audiobooks are one of the most difficult genres to work in.
They require you to tell the story, but you can't be more of a protagonist than the story. In most of them, I listen more to the narrator than to the story itself, which is a problem, as it is the story that has to transcend.
If you are curious, one of the classics to read are Cicely Berry's books "Voice and the actor" and "Freeing the natural voice" from Kristin Linklater.
I love to read aloud for decades, do a bit of different voices per character... And people seem to like my voice when I moderate our daily rounds or do presentations. I just learned that in the last 12 months, so maybe there is a hidden talent that needs to be unveiled.
There are so many genres where a voice actor can do voice acting. Think of anywhere you hear a voice. There is a voice actor behind it.
I don't know what area you live in but if it's something you're interested in, I'd look into where you can take classes.
Sorry, I missed your comment. :\ I'm not a US resident, I live in northern Germany, area of Bremen.
I mentioned to some friends that my wife and I have about 500 Audible books between us on our family plan, that is not counting audio books listened toon Libby from our local town library. I spend easily ten hours a week listening to non-tech audio books a week but only about five hours reading non-tech books.
Several years ago I read James Joyce’s Ulysses and to be honest I didn’t really “get it.” A year later I listened to the audiobook that was narrated by a half dozen skilled actors and the whole thing came alive for me.
I've always been curious about the logistics behind getting an audiobook out.
If the narrator can't tell who is talking, the reader won't be able to either. I think there was only one case where he had to guess, and guessed wrong.
I considered, briefly, narrating my own books, but then after a little research, realized that was a bad idea. It is acting.
I auditioned about 25 actors, both on ACX and on Voices, with a 2-minute section from my book that has all three main characters talking (two male, one female). I really wanted to hear how they did it.
Maxwell Glick (https://www.amazon.com/s?k=Maxwell+Glick&i=audible&ref=dp_by...) won. He's done both of my books now.
Inventing the Future: https://www.amazon.com/Inventing-the-Future-A-Novel/dp/B09NZ...
(On this one, there is a section of Mesa code. I had to tell him exactly what to say, which I modeled on what a CS professor would say if he was reading off the whiteboard.)
The Big Bucks: https://www.amazon.com/Audible-The-Big-Bucks/dp/B0BDBG5VNP/r...
What was interesting for me was, on Voices, some of the actors only gave me 20 seconds, and said they'd do the rest if I wanted! I found this incredible: it's only two minutes, for God's sake. I need to hear how you do the voices.
In their market, some producers will listen to that 20 seconds, say "Nope!" and turn them off. So they're reacting to that. Sad.
Maybe if you don't have people speaking, you can do your own book. Otherwise get a voice actor.
Edit: if anyone would like the text for the samples on the Amazon page, just speak up.
There is a point when listening, it all becomes a tv show in my mind and the ai voices get replaced.
wonder if the same pathways can work for audio, do you use something custom to read for you or something off-the-shelf?
Voice acting is a professional career. I worked with lots of voice talent during my time in entertainments. Like all professional actors they put a lot of effort into developing their range and characters.
I recorded the audio-book of Digital Vegan myself. I am a terrible voice actor but have a good deal of experience with public speaking and radio work. Author-narrated works, especially if technical, have a couple of advantages in that specialist domain knowledge (acronym expansion, pronunciation etc) comes through confidently, and there is some evidence that audiences like author-reads for authenticity.
What I discovered when trying to distribute the work shocked me, and that's why I ended up just putting it on my own site (money was never really the main objective).
The market is dominated by five or six US entities that aggregate smaller distributors and online stores. Buried in the final stanzas of two of five contracts I reviewed were clauses signing over additional rights to machine learning training for the audio recordings plus e-book transcripts. One was actually on a page alone, breaking with the typography and structure of the rest of the contract and following a blank page, so clearly intended to be "snuck-in".
I won't name the companies for legal reasons. But voice actors beware. Read every line of the contracts and if you work for hire for authors it might be worth explicitly excluding permission for machine learning.
Not an issue. You (the author) give them a pronunciation guide, and then you listen to their first draft and make corrections. Possibly even record it yourself for them, as a help.
As for authenticity: you might be right.