Tag Archives: Artificial Intelligence

Creating Auto-Narrated Audiobooks on Google Play

Based on Chuck Litka’s post Adventures in Audiobooks, I decided to go with Google’s audiobook option only. So this post refers exclusively to auto-narrated audiobooks on the Google Play platform.

I have finished setting up and editing one audiobook, which is now live. From that experience, I can make the following observations.

Once you have an ebook live on Google Play Books, it’s surprisingly easy to create an audiobook. That said, I fumbled my way through the process, and my first book went live in an unedited state. I will have to figure out how to keep that from happening with any others I publish, because it is absolutely necessary to listen to the entire book before finalizing it. There are step-by-step instructions in something called the Auto-Narrated Audiobooks Learning Center, but when I started the process, I found myself being hustled through a number of steps without really knowing what was happening. No harm done, however, as far as I know.

On the plus side, once the audiobook was live, it was easy to whip it into shape. I called up the Audiobook Text, which allowed me to both listen to and see the text. I could start and stop playback, make changes in the text, and save them. It’s possible to have more than one version of an audiobook (with different narrator voices, for example), but only one can be live.

There is a large number and variety of narrator voices available, including male and female voices in different age ranges (18-30, 31-45, 45-60, and 60+) with “standard” American, British, or Australian accents. Voices for a few languages other than English are available, but it’s recommended to use these only for texts in those languages. There are no options for English spoken with accents other than those I’ve already mentioned, or with regional accents.

It’s possible to use more than one voice in a book! Theoretically, you could have dialogue in as many voices as you have characters, but I think this would complicate the setup process. So far, I’ve used only one main voice, with a second one to read brief quotations that open a few chapters in my book.

Changes made to the audiobook text are not reflected in the ebook version. One hazard here is I was sorely tempted to improve the text! I don’t know how many instances of the word “that” I was tempted to delete, but I decided to make no changes except those needed to improve or correct the speech. I want the audiobook, ebook, and paperback versions to be essentially the same, stylistic problems and all. However, I did find it helpful to add or delete commas on occasion.

The computer-generated voices sound human, much more so than the rather robotic voice of Word’s text-to-speech feature. Nevertheless, they can’t be expected to represent the full range of emotion that may be found in a work of fiction. Google’s info about auto-narrated audiobooks cautions that they work best for texts that do not require a lot of drama and emotion. Still, I found the voices I selected to be better than adequate. Quite often, the main narrator was spot-on, to the point he seemed to embody the first person narrator of the book.

Quirks and Issues

  • Stress and emphasis does not always fall where it should in a sentence. This can’t be changed by adjusting the speed of the narration, but deleting or adding commas helps sometimes. Still, I admit there are occasions when a word or sentence sounds a little “off.”
  • Sometimes there is an awkward pause between a word or sentence spoken by a character and the dialogue tag such as “he said” or “I asked.” The best solution might be to delete selected dialogue tags, but I resisted the temptation to do this, not wanting my spoken and written texts to diverge. But this is another reason to use fewer dialogue tags!
  • Weirdly, a few names are pronounced quite differently when a possessive is added. For one name, I had to provide a correct pronunciation for possessives because the default was unacceptable.
  • Homographs are common enough that you have listen for instances where the wrong pronunciation pops up. For example, the default pronunciation of the word “read” is the present tense (pronounced “reed”). When the past tense pronunciation (“red”) was needed, I had to intervene. Fortunately it’s easy to fix these; a right click on the word in the text takes you to both versions, and you can listen to them before selecting the correct one. There is an option to change the pronunciation of all instances of a word, or only one.
  • Abbreviations such as Mr. and Dr. are usually pronounced correctly, but I encountered a few situations where “Dr.” came out as “drive,” for some reason. I fixed these by spelling out the word.
  • Uncommon words, place names, or words in other languages may be mispronounced. In such cases, you can insert a different pronunciation by spelling the word differently, speaking it into your computer’s mic, or by using the International Phonetic Alphabet. I actually did that for a few place names; fortunately Wikipedia sometimes provides IPA spellings in its articles, so I was able to reproduce them with good results. You can listen to the new pronunciation before selecting it. This was about the most challenging part of the editing process.

So what do I think of Google’s Auto-Narrated Audiobooks?

I think it’s an excellent option for authors who would not otherwise consider producing audiobook versions of their books. It doesn’t cost anything and produces acceptable results.

There’s no doubt that a competent human reader or voice actor would produce a superior listening experience, but at a cost that’s likely prohibitive for most indie authors. Some may have the talents and equipment to be their own reader, but I suspect those are a minority. The AI-narrated option is available for free to anyone.

A few more considerations:

  • You have to publish your books as ebooks on Google Play before you can create audiobooks. Google requires book files in ePub, not Word. I used Calibre (a free program) to convert a copy of the Word doc I used for the Amazon Kindle version of my book into an ePub, which I then uploaded to Google Play Books. It helped that the Word doc was properly formatted and had a linked table of contents.
  • You need a square cover image for the audiobook, but it looks like the rectangular ebook cover image is squared up automatically with a block of matching colour, so you can get away with that.
  • You need to commit the time needed to listen to your audiobook from start to finish in order to correct any serious or even mildly annoying problems in the finished product. The book I worked with is a fairly hefty tome, which ended up being more than 15 hours of listening time. It took me a solid week to complete, spending 2 to 3 hours each day. (Actually, this reminded me why I prefer reading fiction rather than listening to it.)

I encourage anyone who wants to offer their books in audio format to give this a try. The only cost is your time.

Once I’ve converted one more book to auto-narrated audio format I will write a post on my own blog with more details. That should appear in another week or two.

Featured image from Pexels

AI As a Fiction Writing Tool

by Audrey Driscoll

As an experiment, I wrote a short story I called “The Green Flash” and then presented its basic elements to ChatGPT and asked it to produce a short story incorporating them. You can read the results in this post on my personal blog.

It took the AI a few seconds to produce its 755-word story. It took me the equivalent of 2.5 standard working days, spread over several weeks.

I did not refine the ChatGPT story in any way. But I reduced mine by a third to make it more comparable to the AI version.

I have to admit, the AI’s version of the story, which it titled “Run for Janey,” isn’t bad. It has an arc, it hangs together, it’s even mildly exciting and ends with an “aww” moment. Many would say it’s a perfectly good story.

A few observations:

  • My prompt didn’t say it was the couple’s 50th anniversary. ChatGPT added that detail independently.
  • I did not specify that either of the photographers took photos of anyone, but I guess ChatGPT knows about Chekhov’s gun.
  • ChatGPT seems to have a positive, sentimental outlook, as well as a sense of humour.
  • ChatGPT doesn’t always show rather than tell, and it doesn’t mind using adverbs or stock phrases (i.e., clichés).
  • The green flash is more dramatic in the AI version.

Based on this rather superficial test, I conclude that AI may be a useful tool for fiction writers, especially those under pressure to produce text quickly. It would probably be a good idea for those writers to work on creating effective prompts and refining their inputs.

Have I done that? No. Do I intend to? I’m not sure. I have admitted that crossing the gap from story idea to a first draft is my toughest writing challenge. It’s obvious that AI can do that easily, but I don’t know how much effort would be needed on my part to revise prompts and repeat the AI’s text generating routines before the results would match my intentions for a specific story. It could be I’d rather stick to using my aging brain.

Take the story in my “experiment.” I think “The Green Flash,” written 100% by me, is a better short story than “Run for Janey” by ChatGPT. If I had started by prompting the AI and working with the resulting text, I don’t know what sort of story I would have ended up writing, or how much time and effort would have been required to produce something I thought worthwhile.

Logically, the next experiment would be to prompt an AI with elements from one of the many story ideas in my notebook, and see if I could turn the results into a real story more efficiently than my current process.

The real question is which method of producing written fiction is more congenial and effective for any individual writer. Some may have no interest in engaging with AI; others may be motivated to try different AIs and learn how to use them effectively.

In the latter case, should the human writer credit whatever AI they use as a co-writer? Or is the AI analogous to all the mental inputs a writer has had over their lifetime, from everything they’ve ever experienced, read, and learned, all the movies they’ve watched, all the conversations overheard. Because really, our minds are repositories of all this stuff, and it’s from these wells we draw the ideas for stories and the words in which we embody them. It could be argued that AIs do the same, only more efficiently.

Writers, what do you think of artificial intelligence as a fiction writing tool? What sorts of writing projects have you used it for?

Feature image photo by Audrey Driscoll enhanced with Canva.

Going Straight To The Source

With all of noise about AI and whether creative folks, including writers, should be worried (yes, there was just a tweet asking this very question … again), I decided to go straight to the source.

I asked ChatGPT whether writer should be afraid of artificial intelligence. Here is the answer that I hope settles the debate. 😉

As an AI language model, I can understand why some writers may have concerns about the rise of artificial intelligence (AI) in the field of writing. However, I don’t believe writers should necessarily be afraid of AI. Here are a few points to consider:

  1. Augmentation, not replacement: AI technology can be seen as a tool to enhance and support the writing process, rather than a direct threat to human writers. AI can assist with tasks like grammar and spell checking, generating ideas, providing suggestions, or helping with research. Ultimately, the creative and critical thinking abilities of human writers are still invaluable and difficult to replicate.
  2. New opportunities: AI has the potential to create new opportunities for writers. For example, AI-generated content has been used for various purposes, such as generating news stories, product descriptions, or sports reports. While this may raise concerns about the authenticity of the content, it also opens up new avenues for writers to collaborate with AI systems or explore creative projects that incorporate AI elements.
  3. Uniquely human skills: Writing involves more than just assembling words; it requires emotional depth, unique perspectives, and the ability to connect with readers on a personal level. These are aspects that AI currently struggles to replicate convincingly. Writers can focus on developing these distinctively human skills, which are highly valued in creative and literary works.
  4. Ethical considerations: It is important to be aware of the ethical implications of AI in writing. Issues like plagiarism, copyright infringement, and the responsible use of AI-generated content need to be carefully considered and regulated. By upholding ethical standards, writers can ensure the integrity of their work and maintain trust with their readers.

In summary, rather than being afraid, writers can view AI as a powerful tool that can complement their skills and offer new possibilities. Embracing AI and understanding its limitations and potential ethical concerns can help writers navigate the evolving landscape of writing in the digital age.

It is #3 which is why AI just isn’t much of a risk, at least currently. I believe that AI will always struggle with that. It’s one thing to be able to vacuum up data and predict an answer. It’s a whole other thing to be able to inject human emotions into the output. We shall see.