Making podcasts from Project Gutenberg e-books (2021) | Hacker News Reader

Making podcasts from Project Gutenberg e-books (2021)

charlieharrington.com

63 pointsrevorad45 comments

Voice synthesis is obviously getting better and better, but I've wondered whether anyone has put time into the semi-manual tuning of classic texts for audiobook purposes. Kind of like what Standard Ebooks does for epub typography but for Audio.

Like, can someone edit a file to add dramatic pauses, correct unusual pronunciations and so on to improve the automation? Can someone with a merely average speaking voice record the audio, to provide the human element, but then have that pacing and emphasis applied to a (or several) user controlled automated voices?

Googling it, voice style transfer seems to be a thing, would be interesting to convert the librivox catalog to various celebrities and see how it works.

Azure has really good neural voices with good ssml support. With a sentiment analyses you could probably automate it rather easy.

SSML led me to this EPUB support for enhanced text to speech, so possibly something Standard Ebooks might actually adopt themselves in future.

https://www.w3.org/TR/epub-tts-10/#ssml

It is a nice little project but if anyone actually wants pre-made audio versions of out-of-copyright books then I suggest looking at LibriVox[0]

They have thousands of works read by volunteers (some are very good, some average) and they publish RSS feeds for collections so you can consume them with your podcast app.

For example, The Adventures of Sherlock Holmes[1], a dramatic reading.

[0] https://librivox.org

[1] https://librivox.org/the-adventures-of-sherlock-holmes-versi...

The author links to Librivox at the bottom of the article...

I guess I should have read all the way to the end. But LibriVox is such a great project that I don't regret linking to it again.

See also Magpie Audio for high audio quality recordings of Holmes and others (eg Pride and Prejudice) by professional actors. (They have been put on YouTube by the main reader Greg Wayland for free - though I think he retains copyright)

A well interpreted audio book is really a enjoyable experience. Some books ask to be read aloud.

Many classic radio dramas are still very engaging, and British radio programs on archive.org are also enjoyable. I created a randomizer that plays these through my site:

Classic radio: https://www.locserendipity.com/Radio.html

British radio: https://www.locserendipity.com/RadioShows.html

Also, this text to speech program using the built in TTS API in your browser is useful: https://www.locserendipity.com/TTS.html

As is this project Gutenberg randomizer: https://www.locserendipity.com/Gutenberg.html

Enjoy!

I used to rip the audio of my favourite movies and listen to them like audiobooks. Because I was usally familiar with the visual parts, I could enjoy the dialogue and the music in a different way.

That's a very fun idea. I'll give this a try with my favorite movie. Thanks!

Update: In case someone is curious, I wrote up the steps necessary to do that with an MKV and ffmpeg: https://blog.notmyhostna.me/posts/listening-to-a-movie-as-an...

I frequently 'watch' TV Shows and movies by just listening to them. It's a habit I picked up as a kid since I was blind.

probably not legal, but this sites been around for a long while: https://www.listentoamovie.com/media/index.php

Obviously, I only saw this after slinging obscure FFmpeg commands for half an hour to extract a DTS audio track from a mkv and converting it to a flac file.

Update: In case someone is curious, I wrote up the steps necessary to do that with an MKV and ffmpeg: https://blog.notmyhostna.me/posts/listening-to-a-movie-as-an...

I have always wished Netflix, maybe just for my favorite downloadable titles, allowed for an audio-only option so as not to drain my battery or when I'm out for a walk.

I tried making a podcast of narrated Let's Plays. There are similar YouTube channels if one doesn't mind listening with the screen off or out of sight.

Theres's alternative YouTube clients for Android (and other platforms, apart from the great cli tools mentioned) that help with the audio-only use case that the main apps upsell mechanic breaks accidentally(?) without having to totally reroute all the content.

There's so much spam when you search in Google Podcasts app, I guess as TTS gets cheaper is gonna get worse and worse.

Whether or not it's actually "spam," as TTS gets better it probably gets more tempting to create podcast feeds for articles/blogs because... why not? Even if it's not for monetization purposes.

Right. I hadnt thought of screenplays being read as podcasts before. I know there are some screenplays from the blacklist that I WANT to sit down and read but never will. A podcast version would be great.

I guess it would be ok if podcasts created with TTS included a badge "Automated content" and an option to filter those when you search.

What text-to-speech model is being used? I've seen big improvements in the past few years that could make me actually want to listen to these (since the voices sound more natural now)

Not sure, but I think it's one of the models from https://github.com/mozilla/TTS

That’s the one. There’s a Colab link in the repo so you can try it out easily!

Some days ago I did something similar, I'm using Googles Text To Speech to read Wikipedia and upload it to anchor.fm.

The hardest part was the wrapper for anchor.fm, because of course there is no public API documentation.

The Google API indeed is a bottle neck, because it's only free within a given limit.

https://anchor.fm/rakn

I believe I mentioned this in the article, but I really think scripts or plays would be a great medium for TTS, where you could have separate voices for each character. Could be a neat feature to add to script-writing software.

I find it quite interesting that his final voice sounds exactly like a female speaker standing at a podium in a small auditorium, not someone speaking directly into a microphone like most TTS systems.

There are plenty of applications that read ePubs, PDFs, etc. via TTS, not to mention screen readers. Plus they are more customizable than generating a wav file.

I never understand stuff like this. The idea seems to be to run Gutenberg texts through a text to speech system so you can listen while washing dishes. Doesn't anyone read any more?

Why would it be one or the other ?

There can be people consuming books through reading, audiobooks, or both.

It does not matter much, except for the preference of the consumer : The thing that actually matters is the content of the book, not the channel through which it is consumed.

And even if you think audiobooks are strictly inferior to reading for everyone, I'm sure you'll still agree that an audiobook is strictly superior to no audiobook, which would be the alternative in the case of people listening to audiobooks while driving/doing dishes/etc...

This kind of reading elitism is cringe.

Books can be consumed with eyes or ears. Neither is inherently better or worse. Let others have their preferences without having to endure your elitism.

I do both, but I would argue that text has one major advantage over advantage over audio recordings: it's cheaper, which means that there are many works that don't make the jump from text to audio. Pragmatically, if you restrict what you consume to what you can get in audio, you limit what you can access. Worse still, it isn't even an equal distribution. Popular fiction tend to dominate most audiobook stores, which makes sense due to its broad appeal.

It does seem like this is slowly shifting and good TTS might even make it irrelevant in the future, but for most people consuming via audio only means leaving a lot of material unavailable.

One person has set up a thing = none of the 7 billion people read anymore.

I never understood arguments like that tbf, haha

People experienced literature orally far, far before they ever read it. Listening to stories has been the predominant experience across human history and culture.

Not while washing dishes, no. The amount of people who want to read while washing dishes was always miniscule.

This is the fantastic thing about audio.

You can listen to stuff at the same time as doing some other activity. (e.g. light exercise, or some straightforward task).

I am the target audience for something like this.

I like reading and do read a lot and prefer reading to listening cause for me it's much faster and with better comprehension. But after 10-12 hours of screen time, an audiobook saves my eyes and is actually easier while working out or walking on a treadmill.

I'm frequently doing something similar with TTS on iOS and macOS so this would be an all around improvement.

I can highly recommend the app “voice dream reader”. It’s the first thing I install on a new iPhone. It’s also what I use to read every interesting article that I find on HN. I even scan books simply to be able to read it in the app. Being able to switch between listening and visual reading is amazing. You can also take highlights and export them. Also, no subscription, and works offline with all data stored on your device.

any recommendation for android?

Unfortunately no.

As someone who is neurodivergent, reading is exhausting. I like audiobooks played at 2x speed.

What an odd comment. Can you read while you do the dishes? I cant. Unless I'm listening to an audiobook. I love to read. I do it in lots of formats. A significant portion of my reading comes from listening to audiobooks while doing mindless chores or driving.

Or driving. I spend over an hour driving every day. That's a lot of "reading" time I otherwise wouldn't get without audiobooks.

OP here. I read ALL the time. But I also do dishes. I find it hard to do dishes and read a physical book at the same time. Unless someone else was holding it up for me, slowly flipping the pages. That does sound nice.

> Doesn't anyone read any more?

Socrates faulted writing for weakening the necessity and power of memory, and for allowing the pretense of understanding, rather than true understanding.

https://www.historyofinformation.com/detail.php?id=3439

Reply on news.ycombinator.com