1,000
3.00The market asks about 2026, but I agree that the current state is "NO": https://x.com/GrantSlatton/status/1892488455940428182?t=gSHsMXzn1NIL5vnAeycGSw&s=19
I feel like this this qualifies as YES: https://x.com/elevenlabsio/status/1892593237396763089?s=46
@BraydonDymm I feel like it doesn't, unless people actually want to listen to audiobooks narrated by ai. TTS narrated books existed before this.
@ProjectVictory classic TTS is no where near the quality of eleven labs. Releasing a book-to-AI-generated-audiobook feature for authors on a major platform like Spotify is very much so getting to the core of the question. You would never see an old school TTS published this way, no one would pay for that. The spirit of the question has resolved YES in my mind.
@BraydonDymm Spotify had ai narration for over a year though. What about this announcement is different?
@ProjectVictory “ElevenLabs is one of the most recognizable AI voice providers on the market, however, which could lead to a surge in synthetically voiced audiobooks on Spotify’s platform.”
“Authors can use ElevenLabs to narrate their audiobooks in 29 languages, with a wide selection of synthetic voices to choose from.”
@ProjectVictory sounds fair to me. E.g. we could get strong reasons for YES or NO respectively if the functionality gets celebrated as a big success and copied by competitors, or quietly sunset without further adaption.
@HenriThunberg that’s not the question though, the question is not “will AI-generated audiobooks become more popular than human ones”. Whether or not the functionality is celebrated, copied, leads to a surge, or whatever, the fact of the matter is plain and clear: human narration for audiobooks is mostly unnecessary.
@BraydonDymm well what you sent it "AI company claims it obsoleted human narrators". Unless you have a better objective measure of book narration quality, I'd say we judge it based on adoption and public opinion.
@ProjectVictory the "mostly" in the title make it clear that this market run more by vibes than by metrics. But I am not in a hurry to resolve it.
@FranklinBaldo I just tested the reader app , generate a podcast, and you can perceive that the narration is soulless. Notebooklm much better.
I've listened to probably a hundred hours of AskWho Casts AI narrations powered by ElevenLabs, most notably Significant Digits, and it's been excellent, much better than Type III Audio. I don't think I could differentiate it from a human voice except that the voice doesn't change much depending on the content, and occasionally it has been inconsistent in the pronunciation of characters with made up names or has an incorrect pronunciation of an abbreviation. The premium I would pay for a human narrator would be very small.
AskWho did an interview a year ago, notably he said that it cost $12-15 per hour of audio and that it still required a human to annotate the text with which lines should be spoken in which voices because LLMs weren't reliable enough at that yet, though maybe this has changed in the last year. This only affects fiction though, not nonfiction which uses one voice.
I wonder if the reader app is using one of their cheaper, lower quality models. Also you typically see more emotional range in the voice in a podcast than in an audiobook, so it's probably not well suited to the conversational podcast format.
@ahalekelly I listened to hours of these as well, it's great. You certainly need to listen carefully to notice it's AI, but notebooklm I sincerely wouldn't recognize as AI if I didn't know theirs voices.
@FranklinBaldo I think the key point here is that, regardless of whether we are personally okay with or even prefer AI narration, it's hard to make the case that this is true of most people unless (i) we see blinded, preregistered, and rigorous experiments showing this, or (ii) we start seeing mainstream audiobook publishing turn to AI narration.
As mentioned, there are other factors for the industry like inertial forces of human narrators and people whose jobs depend on hiring, recording, etc., but it's hard to make the case without at least some substantial drop in human narration.
I say this as someone likely in the 99th percentile of hours spent listening to TTS.
How do you plan to judge this market?
This feels like “another AI fan creates a market and precommits to resolving it YES if they remain an AI fan,” which isn’t very interesting
correct. English only, as @Bayesian pointed out. (clarification added due to ambiguous manifold ui)