Blind and Visually Impaired Community
↑
2
↓
Feedback on ebook creation needed (self.Blind)
submitted by Simons_Mith
I am researching improving the quality of 'robot-read ebooks' and would like to pick the brains of anyone who uses them.
Do many people here listen to audio ebooks? There are two common approaches - either an actor in a studio reads it, which is expensive, or you can use a text-to-speech tool to play an ordinary EPUB or MOBI format ebook via a robot voice. Which in theory means you can listen to any ebook, but only in a robot voice. The vast majority of people use the former, but I'm trying to find anyone, blind, sighted, or ebook developer who listens to/creates ebooks via the second method. If that's you, please can you contact me? (Or if it's someone you know, please pass this on. I'm having great difficulty getting answers to my questions. Even the RNIB haven't got back to me.)
It would be especially helpful if you are familiar with the W3C's CSS Aural Reference: https://www.w3schools.com/cssref/css_ref_aural.asp but anyone who listens to robo-books would be great. Thanks!
Do many people here listen to audio ebooks? There are two common approaches - either an actor in a studio reads it, which is expensive, or you can use a text-to-speech tool to play an ordinary EPUB or MOBI format ebook via a robot voice. Which in theory means you can listen to any ebook, but only in a robot voice. The vast majority of people use the former, but I'm trying to find anyone, blind, sighted, or ebook developer who listens to/creates ebooks via the second method. If that's you, please can you contact me? (Or if it's someone you know, please pass this on. I'm having great difficulty getting answers to my questions. Even the RNIB haven't got back to me.)
It would be especially helpful if you are familiar with the W3C's CSS Aural Reference: https://www.w3schools.com/cssref/css_ref_aural.asp but anyone who listens to robo-books would be great. Thanks!
fastfinge 5 points 6y ago
I read ebooks almost exclusively with my screen-reader and the "robot voice". I find actors too slow, generally. Plus I'm so used to the Robot voice that I no longer hear it, and can just imagine any voice I want. Unfortunately, though, the only thing I know about aural CSS is that nothing supports it. Seriously, nothing. Not NVDA, Not JAWS, not VoiceOver, etc. If you can point me to anything at all that supports it, I'll be shocked.
Simons_Mith [OP] 2 points 6y ago
Aw, crap, unless anyone else knows different that pretty much answers all my questions in one go. :-( Thank you though.
fastfinge 3 points 6y ago
What are you trying to do? It's possible that you're asking the wrong questions. :-) For example, if you're trying to indicate bold text by speaking it in a different voice, most screen readers can indicate bold text already, in ways that the user can configure to work best for them.
Simons_Mith [OP] 2 points 6y ago
I work for a company that produces ebooks, including non-fiction and drama. When an ebook is read with the robot voice, the voice will presumably be the same for the whole book. Headings and sub-headings may get minor customisations, like pauses after they're read, but our books already do those bits correctly.
I was looking at CSS as a possible way to define different voices for different characters, and for stage directions. But if I did that, would anybody like it? Could it actually be worse than having the whole play read in one single voice? I don't want to over-customise this stuff if most robo-voice listeners would dislike the results.
Say there was a child and an adult character, and I made a childlike voice by using the CSS (that nobody supports) to raise the voice pitch and speed, and slowed the adult voice down slightly. But if you've already configured your reader to read the default voice at the fastest intelligible speed, the child-voice might cross over into unintelligibility. So while there's things I _could_ customise, (speed in this case), I need to canvas the preferences of actual users on how much customising I _should_ do.
And if you're used to a monotone generic robot voice, you might even find an increased variety of robo-voices a distraction, because they'll still remain unavoidably robot-like. On the other hand if you've got a play with a dozen characters, having them all read in the same voice could also be confusing. So, what's the best compromise? And I bet it's different again for a plain novel or a piece of non-fiction.
If I did customise the voices, how readily distinguishable would they be? Can I use CSS to make only 4 to 6 distinctive voices, or could I make 20? Are there any libraries of CSS voice settings I can refer to? There's lots of customising you can do - how much difference do all those settings actually make in practice? Are there any examples of ebooks with custom voices that I could study?
It seems like the answers to most of these questions are basically 'nothing supports that' or 'no'. So for now I'd probably do better to ask specifically about CSS support before worrying about particular implementation details. I need better support first, before I can implement the customisation I'd like to do.
I was looking at CSS as a possible way to define different voices for different characters, and for stage directions. But if I did that, would anybody like it? Could it actually be worse than having the whole play read in one single voice? I don't want to over-customise this stuff if most robo-voice listeners would dislike the results.
Say there was a child and an adult character, and I made a childlike voice by using the CSS (that nobody supports) to raise the voice pitch and speed, and slowed the adult voice down slightly. But if you've already configured your reader to read the default voice at the fastest intelligible speed, the child-voice might cross over into unintelligibility. So while there's things I _could_ customise, (speed in this case), I need to canvas the preferences of actual users on how much customising I _should_ do.
And if you're used to a monotone generic robot voice, you might even find an increased variety of robo-voices a distraction, because they'll still remain unavoidably robot-like. On the other hand if you've got a play with a dozen characters, having them all read in the same voice could also be confusing. So, what's the best compromise? And I bet it's different again for a plain novel or a piece of non-fiction.
If I did customise the voices, how readily distinguishable would they be? Can I use CSS to make only 4 to 6 distinctive voices, or could I make 20? Are there any libraries of CSS voice settings I can refer to? There's lots of customising you can do - how much difference do all those settings actually make in practice? Are there any examples of ebooks with custom voices that I could study?
It seems like the answers to most of these questions are basically 'nothing supports that' or 'no'. So for now I'd probably do better to ask specifically about CSS support before worrying about particular implementation details. I need better support first, before I can implement the customisation I'd like to do.
fastfinge 3 points 6y ago
One of the reasons nobody supports the CSS is because TTS voices just aren't like fonts. There is no standard library of available voices that you can depend on existing, and no standard way of describing the voices that do exist. For example, on my system, the SAPI voices I've got are Cereproc Catherine, that tags itself as "female, north eastern american, adult" and a voice called Tamar that just tags itself as "young adult, female". If you asked for an adult male voice, I've got nothing installed to match that.
Secondly, even if there was, I would resent anyone randomly changing my TTS settings. I have the voice I want, set at the speed I want. I'm so used to it that it fades into the background.
If you're marking headings, formatting, paragraphs, etc correctly, that's all you need to, or should, do. Do your ebooks have images? Have they been described correctly? What about tables? Are they marked up properly? Those things are way more important than worrying about aural CSS.
Secondly, even if there was, I would resent anyone randomly changing my TTS settings. I have the voice I want, set at the speed I want. I'm so used to it that it fades into the background.
If you're marking headings, formatting, paragraphs, etc correctly, that's all you need to, or should, do. Do your ebooks have images? Have they been described correctly? What about tables? Are they marked up properly? Those things are way more important than worrying about aural CSS.
-shacklebolt- 2 points 6y ago
> But if I did that, would anybody like it? Could it actually be worse than having the whole play read in one single voice?
As another blind user, I have to echo "yes it would be worse" at least as far as many blind users are concerned.
I personally dislike most human-read audiobooks (not fast enough and I hate it when they try to put on character voices or narrate animatedly), and like /u/fastfinge/ my preferred "robot voices" are so normal to me that they can take on whatever voice needed in my head for the story. Consistency makes for easy listening, maintaining my reading speed, etc.
As another blind user, I have to echo "yes it would be worse" at least as far as many blind users are concerned.
I personally dislike most human-read audiobooks (not fast enough and I hate it when they try to put on character voices or narrate animatedly), and like /u/fastfinge/ my preferred "robot voices" are so normal to me that they can take on whatever voice needed in my head for the story. Consistency makes for easy listening, maintaining my reading speed, etc.
This nonprofit website is run by volunteers.
Please contribute if you can. Thank you!
Our mission is to provide everyone with access to large-
scale community websites for the good of humanity.
Without ads, without tracking, without greed.
scale community websites for the good of humanity.
Without ads, without tracking, without greed.
©2023 HumbleCat Inc • HumbleCat is a 501(c)3 nonprofit based in Michigan, USA.