Hacker News Reader

Comment by dmitriid | Hacker News Reader

”A few hours with a few people” is

- often multiple hours in the studio because people make mistakes, they get tired, they want to re-do passages etc.

- during recording you want at the very least a sound technician, the voice talent, and a director/author/producer to listen in and give direction/corrections/adjustments on the reading

- after recording you need at least an editor, a director/producer to make sure the recording is cut correctly.

And then off to the presses (that is, converting RAW audio to whatever digital formats).

A designer may also be involved if the book requires a different cover.

Disclosure: I work at Storytel though not anywhere near (other than physically) audiobook recordings.

View on HN

People do Librivox recordings free of charge as volunteers. Those recordings are just as good as Audible recordings. I myself, with rudimentary gear and a free DAW, can create an Audible level recording for the mere cost of my time. I'm going to have to totally reject your list of expenses. Sounds like a bunch of superfluous positions and bloat.

Isn't this the equivalent of "I can do Dropbox in one weekend"?

Or, more recently, "someone with one week of JavaScript experience can replicate Airbnb."

If you can't tell the difference between the average LibriVox recording and the average commercial audio book, there is no way you can produce an Audible level of recording.

Even if half the volunteering recorders and their gear were as good as those of the commercial offerings, just sitting and selecting those that are not good enough for publishing and those that require some re-recording is a massive extra work.

What is so special about Audible recordings that is unnattainable by Librivox? You mean sound quality? You mean some holy grail vocal mics? Or vocal talent? I don't get it.

Mostly talent, production, and editing.

I've listened to a few Librevox recordings. They're usually (though not always) tolerable. Occaisionlly good, but that's exceeding rare.

Good audiobooks -- and a friend listens to many -- are far more often vastly superior. Even then, a poor reader is exceedingly grating.

Libravox may improve with time. History of volunteer efforts has been compeling. But not yet.

Its not that high quality is unobtainable, it's that nobody at librivox seems to try to attain it. Sound quality is definitely uneven at librivox, ranging from poor to middling, and passages composed of multiple tales are rarely mixed together smoothly. The vocal talent is mostly adequate (I certainly find plenty of narrators I dislike on audible) but it also seems sort of unsupervised -- narrators who encounter unfamiliar words will mispronounce them. I honestly think they'd get more even results with TTS now that WaveNet is so good.

I'm impressed by what librivox has achieved with just volunteers, but the production values are nowhere near as high as most books on audible.

Ratings could be a partial fix for quality control but they have been opposing them for years (because they want to keep it fun and motivating for volunteer readers..): http://piratelibrary.com/2010/05/on-the-absence-of-ratings-a... (ratings are now available in some LibriVox apps though)

IMHO they are just doing injustice to a number of good readers in their catalogue who put real effort into their recordings, but get buried under those one can barely listen to.

> Those recordings are just as good as Audible recordings.

No. They are not. I tried listening to Don Quixote Librivox. While a couple of readers are tolerable, several are terrible, and none are close to professional voice actor level. Which makes it really hard to get through. Check out the comments here too if you don't want to subject yourself to listening to the reading itself:

https://www.youtube.com/watch?v=fLTTJo0E1ag

Audible recordings in general are vastly superior.

Partly true. I've listened to a handful of excellent readers on LibriVox. For audiobooks I don't care if audio quality is sub-par as long as I understand every word and the reader is good.

There were even these incidents with people ripping the best recordings off LibriVox and selling them on Audible where they ended up achieving high ratings..

However, bad LibriVox experiences were one reason why I didn't truly "get" how superior audiobooks can be for years. Enduring their endless copyright and chapter announcements also hampered the experience a bit.

LibriVox derivatives have now fixed the problem of finding good readers in some of the apps with simple user ratings. I believe a similar project based on WaveNet will make much of LibriVox obsolete soon by being better than many readers there, which saddens me a bit since in principle LibriVox is a good project that simply promoted free sharing of information and knowledge.

Both AI and most of LibriVox have a lot to learn to even come close to the quality of the professional Gert Westphal reading of "The Magic Mountain", or to anything Stephen Fry has read.

I listen to a lot of librivox books. There are some talented volunteers, but there are many that are just not that good.

Plus all their text is on the public domain.

Reply on news.ycombinator.com