.

Thursday, May 25, 2023

Producing Audiobooks Using Artificial Intelligence: beta mode. Louis Shalako.

She can be trained...




Louis Shalako



A few notes on producing audiobooks, narrated by AI, on Google Play.

This feature is free while it’s in beta mode. I was uploading my new novel, A Stranger In Paris, an Inspector Gilles Maintenon Mystery #9.

It took three winters to write that thing, and I won’t say I was in a hurry—

As soon as I signed into the website, I saw the big blue button. Create an auto-narrated audiobook, click here.

Hmn. Interesting, I thought—

We will check that out later.

It clicked in after a while that they might be charging money for that, as soon as it goes into regular mode.

With 150 titles on Google Play, full-length novels, novellas, collections, short stories, this represents something of an opportunity. And early adopters often reap the greatest benefit from new technologies.

***

You have to read around in there, you sort of have to experiment. And yet, this is my book—but you have to take some risks. The ebooks are well formatted. The POD interiors are good, although I have yet to crack the code in terms of getting a cover uploaded on Lulu (dot) com, for example.

Okay, so your cover should be square, (not a regular portrait shot) and on the first couple of uploads, I went to the trouble of modifying the original ebook cover. Yet in the trouble-shooting phase, I read that Google Play will ‘letterbox’ your original image. Why make it square when you can just use the original marketing image. (Downloaded in full size from Smashwords. - ed.)

Also.

(If you’re just starting off and you have a small number of titles, you may want to do the work, but also, keep on going in the same manner. – ed.)

He’s right. Consistency is good. #Louis

***

Front matter. This tends to be repetitive, with a title page, a copyright page, your name in there twice or three times over two pages and a small number of words. I’m trimming that down to the minimum, also the author needs a unique ISBN number for each additional format, for every single product. You can get them from Collections Canada ISBN service, free to any Canadian citizen.

Okay. My book has chapters, my shorter works have scenes, acts, or whatever. In an Epub ebook, in order to get maximum distribution through a platform like Smashwords, you need internal navigation. You have a Table of Contents, which is in text, followed by a bunch of links to individual chapters. All of this is simply ignored. The links disappear, but you have to delete Table of Contents, which is merely text. Creating a credible audiobook, even with the help of the tool of AI, turns out to take some effort after all…you have to listen to every word. An audiobook can go six hours of listening time, and fixing things drags that out even longer. In my case, it would have been smarter to start with the absolutely shortest stories on the list.

***

Okay, so front matter, and end matter, that can be deleted or rewritten. For example, my bio also has a link to my blog. The bio remains, in the audiobook interface on Google Play Partner Centre, but the link is gone, what remains > Louis Shalako < (exactly as it appears. – ed.) remains. Again, this should be removed.

I should go back and check and see if the machine says ‘The End’ at the end of the book…

***

Editing. As for editing the actual text. "Hmn." will render as H-M-N, heh-heh-heh suffers a similar fate. Hyphenated words, for the most part, have a funny cadence and so now I'm taking out hyphens, one at a time...Chapters are deleted, so I put the cursor on the first line of text and bump that down, go back up, type in Chapter One, etc., then the machine will read it properly. You can adjust speed, you can pick voices, which are kind of limited in English. There are several other supported languages, and this is when it becomes all the more impressive, in terms of automation and AI.

So, having learned this much, it seems one should go back and review the first few audiobooks I have produced. That being said, the Google Play AI reads the text fairly well. The learning curve, as always, is fairly steep in the early stages.

I've tried changing the pronunciation of my name, which is not good. I have been unsuccessful.

I’ve learned a hell of a lot in the last two weeks, one could safely say. The machine has its limitations, but then so do I. So do we all.

Louis will crack the code.

#experimental

I had one book that just would not go yesterday, I tried again today and it still won’t go. I have no idea of what is wrong, and yet I am reluctant to chat with Google. Yet I’ve uploaded a dozen, converted them, and hopefully they are relatively good products.

The Bio.

It seems like you can update your bio, yet it’s not immediately evident on your first pass. I’ve published, and seen the product go live, and yet I seem to have skipped past the part where you select a genre, and put in the author’s bio.

You can’t get the book to go live without setting a price, yet you can mess around on the page for some time. A book should go live within 24 hours, my typical experience is ten or fifeen minutes at most.

Analysis.

Okay, so in beta mode, staff are working to perfect the thing and I’m working to perfect my thing. They must be learning a lot as well, and as long as it’s free, why not jam up as many titles as possible while the going is good. I can always go back later and fix things, I routinely do quality control and yet that is one big project for another day.

The full (hyphen) length books would be labour intensive, to say the least.

In a recent comment on Facebook, someone called AI ‘plagiarism software’. I cannot plagiarize myself. I own all copyright on products created using the tools otherwise known as AI.

AI would have been trained on the same open-source resources you and I also have access to, namely books, news sources, encyclopedias, as well as large spoken-word sources of a similar nature. They train it, (you and I train it), using music, and pictures and video and any number of other sources.

In that sense, it is ‘learning’, or being trained, and who knows where all of that is going to go. As for myself, I figure I can defeat any system by using its own rules against it.

I will try and stay ahead of it.


END

 

Formatting A Stranger In Paris (ebook).

Experimental: Artificial Intelligence.

 

Poor old Louis has books, stories and yes, ladies and gentlemen, audiobooks on Google Play.

 

See his works on ArtPal.

 

Thank you for reading.