Unnatural audibles
« previous post | next post »
I'm so far behind the times with gadgets and trinkets and services that I have never listened to a single audiobook, and I had never even heard of Audible until yesterday when Gene Hill told me that his wife, Marri, listens to tons of Audibles because she writes reviews, and as a result they give her lots of free stories to read. Of late, the publishers of Audibles are using narration by AI.
No way to overemphasize the importance of the quality of narration in an Audible. Marri most often prefers to have the author do the narration. Only the author knows how to express the precise emotional quality to a line. Or deliver the right touch of sarcasm.
In a tech work, precision of words and grammar are a must and there is no requirement for emotional expression. But in the art of storytelling, it is required to stimulate the reader's empathy, You can't share the trials or triumphs of a protagonist without the energy of all those emotional stimulants.
Marri described the AI narrations as like listening to a six-year-old read a story in a foreign language. In a monotonous drone that would cure insomnia.
No doubt the AI narration will improve over time, but I'm almost certain they will never be able to meet the demands of most listeners for emotive expression.
So says Gene, and I'm inclined to believe him, though I wonder how far AI narrators can go in capturing the nuances of a literary text. This should be another fun part of the AI humanization ride.
Selected readings
- "Sentence punctuation to indicate slowed speech rate" (5/25/08)
- "Storytelling: What’s it good for?" (1/14/19) — "stories are a shortcut to our emotions"; storytelling and science
Terry K. said,
March 8, 2024 @ 5:49 pm
Thankfully, Audible has previews to listen to before you buy, usually about 5 minutes worth. I haven't run into AI narration yet, but even with human readers, you might find one where you don't like the narration.
mg said,
March 8, 2024 @ 8:47 pm
I don't usually do audiobooks, but my friends who do have favorite readers. Authors are sometimes good (Mary Robinette Kowal is an excellent voice actress with a theater background, as well as SFF author) but some are just not good readers. There are some authors who use the same voice actor for entire series (MRK, for example, does Seanan McGuire's October Daye books). Years ago, my kid and I loved listening to the All Creatures Great and Small series while driving cross-country, because the reader was the actor who'd played author James Herriott in the TV series.
I'm horrified at the thought of fiction or poetry narrated by AI. It's probably fine for textbooks and the like.
Anthony said,
March 9, 2024 @ 11:18 am
In the 1970s I had job with the evening shift at Recording for the Blind. The volunteers had a choice of reading matter, and textbooks were rarely their choice. When not splicing reel-to-reel tape or showing people what to do, I read the textbooks myself. When issued on cassette, the tapes often got sped up via a control on the tape players issued to our clients.
Richard Hershberger said,
March 9, 2024 @ 12:00 pm
@mg: Sure, but get the actor who played Tristan and you also have an actor who played the Doctor!
Victor Mair said,
March 9, 2024 @ 12:29 pm
AI-Generated Marilyn Monroe Chatbot Can Hold an Extended Conversation With ‘Realistic Emotions’ and Expressions, Company Claims
Variety
Todd Spangler
March 8, 2024 at 12:05 PM
Marilyn Monroe, who died 62 years ago, has been reincarnated as a “hyper-real” AI-generated digital avatar that lets fans engage in a conversation with the late actor — who can answer questions “in Marilyn’s signature voice and style,” according the company behind the project.
AI technology firm Soul Machines unveiled Digital Marilyn, an interactive experience in which she “shares anecdotes and even delivers personalized greetings, creating an unforgettable experience for Marilyn fans.”
Jarek Weckwerth said,
March 9, 2024 @ 3:39 pm
At the moment, Elevenlabs advertise themselves as leaders in "expression-rich" Text-to-Speech. While I'm not totally sold, the results tend to be way better than what was available even two or three years ago. The main weakness, I think, is that the "expressive" aspect tends to be somewhat random. I'm not a fan of audiobooks, so I can't judge if they are good enough for that, but that's one of their selling points. Give them a try and judge for yourself.
They also do voice cloning. Haven't tried that but their multilingual voices suggest it may be convincing.
Victor Mair said,
March 9, 2024 @ 4:20 pm
From Gene Hill (by "Audibles", I think he means it in the generic sense, to include audiobooks, not the proprietary sense):
I have been listening to Audibles since 1990. That was the year I went blind (detached retinas). I was forced to lie in a prone position for 6 weeks at a time. Couldn't lift my head up to watch TV. Audibles were a god send.
I listened to every tape our local library had. Went through about seven operations and each time had to go through the same healing process. At some point in the mid-nineties, A son-in-law gave me his old computer. Mine was out of date and only used for work. So once online I discovered Audibles. And other sites like The Gutenberg Project. Marri fell in love with Audibles and still goes to sleep every night listening to them. I found it a profound variation to reading. I am a very slow reader because I may read a sentence several times to relate the subject to other things I know or have read. And often go back and read a whole chapter to myself in the manner I presume the author may have intended. Audibles is a different experience altogether. They took me back to sixth grade to a teacher who would read to the class. She had a beautiful voice, and she led me to appreciate poetry. She introduced me to Rudyard Kipling. And encouraged me to read it with the slurry Cockney accent of an old British Tommy. "On thu road to Mandalay, whur thu floyin fishes plays. Aud thu Soun cums up like thunder, outher China cross thu bay."
We wore out several I-pods. One died with about a hundred unread books onboard. But still, the online experience fills our lives with exposure to all sorts of sensual and intellectual stimulation to broaden our consciousness. And after all, that's all we really have.