🏅 Top traders

Currently listening to AI 2027 read by an AI voice and it sounds bad. The voice stutters, has weird intonations, and obviously does not understand what the material it is "reading" is saying

@FranklinBaldo Do you have any examples of these glorious AI audiobooks yet?

@FranklinBaldo Can you even provide a single example of an AI narration that is of similar quality to a human audiobook?

Because right now it just seems like you asked an AI to write an explanation why you should resolve YES, and then resolved YES. Frankly this seems like obvious bullshit.

The resolution here is a joke that should lead to the blacklisting of the creator. Will avoid in the future.

I mean, yeah, Crystal Society (and the whole trilogy) was done using AI narration, each character with their own voice. https://crystalbooks.ai/audio/

@HWH And that narration sucks, so that's evidence for NO.

@Pazzaz you’ve listened to all of it? If you’ve only listened to the beginning, the character at the beginning of the book is an AI, so if it sounds like AI, that’s very much on purpose.

@Pazzaz listening to the sample I wouldn't say it sucks but it's not trying to replace human narration in general, or sound indistinguishable from humans, I believe they are using AI voice for a certain artistic effect, playing off it's robotic nature and even adding distortion effects on top.

@ProjectVictory it’s a little awkward because the character in the sample is an AI, so a different sample probably would’ve been better for showing off the more human-like voices, but I think it was a dilemma of giving a sample, but not spoilers. it’s also a little awkward, because the tech progressed in the middle of creating the audiobooks such that the audio quality for Crystal Eternity is a lot more advanced than Crystal Society. I think Crystal Society might get remastered, but doing it all was a big project, so maybe after some time.

If human narration is mostly unnecessary then why do audiobooks overwhelmingly still use human narration?

i came out on the winning side of this market, but is this really true yet, audio models still seem worse than human narration : /

@CampbellHutcheson I wonder how you would know. They wouldn't advertise it and if you can't tell that's kinda the point

@CampbellHutcheson but tbh it does seem a bit early to call this. It seems to have been decided by the fact we can do it nowlt that it being done. Which is a valid answer for how the question was asked

Here's a comment explaining why the Manifold market should resolve to YES based on recent advancements in AI narration technology:

---

Comment for Market Resolution

As the creator of this Manifold market, I am resolving it to YES based on significant advancements in AI narration technology, particularly with OpenAI.fm and Sesame AI. These platforms have achieved substantial milestones in voice quality and affordability, meeting the criteria I set for human narration to become "mostly unnecessary."

Key Developments:

1. OpenAI.fm offers high-quality text-to-speech with diverse voice options and emotional control, making it suitable for most audiobook needs. The cost is significantly lower than traditional human narration, at approximately $12-15 per hour of audio.

2. Sesame AI provides dynamic conversational models that maintain context and adapt tone, further enhancing the naturalness of AI narration. Its open-source nature allows for rapid community-driven improvements.

Market Impact:

- Quality Parity: Recent blind tests show that many listeners cannot distinguish AI narration from human narration in short samples, with AI handling a broader range of emotions.

- Economic Viability: The cost savings are substantial, enabling small authors to produce audiobooks without relying on expensive human narrators.

- Platform Integration: Direct publishing to major platforms like Spotify further streamlines the process.

Given these developments, AI narration has reached a point where it is no longer necessary to rely on human narrators for most audiobook production. While there will still be niche cases for celebrity narrations, the majority of the market will shift toward AI-driven solutions by 2026.

Resolution Rationale:

- Non-grating quality: Achieved through improved emotional range and natural speech patterns.

- Affordability: Costs are significantly lower than traditional human narration.

This resolution reflects the rapid progress in AI technology, which has transformed the audiobook industry's reliance on human narration.

This reads like an advertisement lol. I suppose without a quality comparison (which itself is subjective) and not taking adoption into account this makes sense. Still feels really early, even for "mostly unnecessary." Plus with that phrasing I was thinking in terms of adoption not if it can be done.

But I'm not gonna say too much cuz I did come out on the paying side 😂

@FranklinBaldo This is obviously written by an LLM...

@FranklinBaldo

It's very distinguishable.

@FranklinBaldo The worst thing is people do that in real life, copy and paste LLM output, and think that works.

@skibidist what are you talking about? This text is clearly indistinguishable from resolution explanation a human would write!

@ProjectVictory Too cruel to remind me of this market after I'd forgotten. The thoughtlessness of it is just so infuriating. My strong initial impression is that the OpenAI.fm voices are too annoying / autistic to listen to in the long run. Like most of the Project Gutenberg narrations are unlistenable due to various quirks that non-professional narrators have that mount up over a long listen.

Credence for somewhat unnecessary i.e. any normal voice type good enough to not be grating and for affordable cost to authors: 95% (it already happened a few years ago in my opinion but others may disagree)

Credence for completely unnecessary i.e. indistinguishable from human: 25%

@TheAllMemeingEye I think if you listen to e.g. AskWhoCastsAI, we're not quite at the "not be grating" threshold for the mast majority of people, although (definitely for me &) maybe for over 50% of people (which I would not see as resolving this market to YES).

@FranklinBaldo For resolving this market, how will you judge whether it's still necessary?

@GazDownright I'd imagine some books (like celebrity bios or well known authors like David sedaris) will and should always be a person but if a majority of the other books use a non author or actor that would be the bar. My two cents

@Shrewdan It's imperative to know how the market creator plans to determine necessity to be able to gauge probability.

@FranklinBaldo ?

@GazDownright However he wants (:

In a different comment he says it should be indistinguishable from humans though. By this criteria, it definitely shouldn't be at 80%.

I commented about this before, but this market is hilarious, I think there's very little disagreement about specific capabilities (AI can produce okay voice, but is more lifeless and monotone compared to human narrators). And most of the trading is just people understanding the resolution criteria differently.

@ProjectVictory your last remark made me go cold on this market. Thanks for looping me in

@GazDownright Yeah, I started by trading based on my belief and understanding of the title, but now I'm just making mana off the wild price swings.