Tuesday, September 19, 2023

The Intangible Man Audiobook

Started work on the audiobook version of The Intangible Man & Other Strange Tales today. There are a few technical requirements. The biggest hurdle is probably the file sizes for so much audio, perhaps 1Gb of data per title. More than a complete music album.

Also required is a new 2400x2400 pixel cover, which is far bigger than the book requires. I've decided to use the audio album standard, which is currently 3000x3000 pixels. It's somewhat odd that these similar media have different standards; an audiobook is little different from a music album in appearance and how they are sold. 3000 pixels (plus bleed) errs on the side of future compatibility. Here's a look at the new cover (the cover so far, that is...):

I had to make a surprisingly large amount of changes for this.

The volume requirements are relatively simple, though far quieter than the standard for music (which is measured in LUFS, and more like a loudness limit than a recommendation). Book audio is limited to an -18dB average, which is about 75% of a regular peak, and at least 3dB of headroom is required at all times.

I read about 10 stories before my voice became too tired for more. I'll split the process into recording first, then editing and processing. The editing will take longer than recording and it would be slow and distracting to dart between one job and another. Reading is a lot like acting, bringing the story alive; it's not just a matter of clarity.

AI readers (text to speech) are forbidden, perhaps for job-protection reasons more than anything else, which is a unusual stance for the commercial exploitation of a technology. YouTube, for example, would never ban AI generated videos (most of the screen-time for today's Hollywood films would count!). Twitter would never ban automated text, or Amazon AI-written books, or the sale of 3D-printed, computer designed objects. Human customers would almost always prefer human readers to listen to, but if offered a worse-quality voice for less money, some people would pick that, and history indicates that cheap, poor-quality, and disposable generally wins in commerce, however lamentable that fact is. I expect that banning 'robot' voices will be a losing battle, but I also know that no AI voice, today, could do the job as well as a human, and that readers/listeners would prefer a real person - expecially the voice of the author. One day, AI actors and AI voice artists will be as 'good' as humans, though, like humans, each will be unique and imperfect. Perhaps an AI Charles Dickens or AI Shakespeare will be our best hope of hearing their words in 'their' voice...

This work is, for me, an experiment. I'll see how it goes. Some books lend themselves more to audio than others. My How To Organise Your Computer Files, my second biggest seller, won't work as audio because its a complex reference book that needs information to hand - it works best as a paper book. Stories work, but wordy literature can be hard to listen to; that's best read at your own pace. To absorb, and to see and appreciate the language.