Speech to text

[OMG]

Maybe wrong place to ask but here goes.

She wants to write a children's book, has the story already, and needs an efficient way to get it to text.

Typing skills are ok but is far more confident speaking the story, "it flows better".

I have Macs for her to use. Also a machine that can record to CD.

Suggestions?

Any advice on story to book in sale form sincerely welcome.

The main character is a mouse with no doubts as to it's gender.

[blackhawkxx]

PM our author Ironmike.  I bet he knows.

Edited December 12, 2023 by blackhawkxx

[SwampNut]

6 hours ago, OMG said:

Maybe wrong place to ask but here goes.

She wants to write a children's book, has the story already, and needs an efficient way to get it to text.

Typing skills are ok but is far more confident speaking the story, "it flows better".

I have Macs for her to use. Also a machine that can record to CD.

Suggestions?

Any advice on story to book in sale form sincerely welcome.

The main character is a mouse with no doubts as to it's gender.

 

Either you're not aware of Apple's built in dictation, or you/she think it's not good...?  Dragon has always been, and currently is the absolutely most accurate, most professional product.  It seems like every single law office uses it.  You can (should) further enhance this with a high quality headset to minimize noise and maximize the accuracy.

 

And then, there are human-driven correction services that can further fix up the text.  They typically employ people in the third world for a few pennies an hour, and work decently well.

 

[XXitanium]

My wife uses her iPhone to take dictation of business documents. 

 

If you help you SO, that makes you handy, man.

 

[ironmike]

Among authors it's referred to as voice to text; and to be honest it's not held in high esteem. Most use it as a temporary short-cut to record ideas or sudden inspiration.There are a host of similar voice to text programs (Google Docs, Apple, Gboard, Dragon Speaking, etc.) but they almost all suffer the same shortcomings. Problem areas include grammar, punctuation, capitalization, homonyms, translated language terms, etc. At some point the text must be edited and retyped in an acceptable draft/manuscript format.

 

I don't want to discourage any aspiring author--on the contrary, I encourage the attempt to give life to the idea and craft the tale that wants to be told. But the reality is that it must appear in an acceptable written format to have any hope of publication, be that trade or self-published.   

 

Curiously, the current trend is text to voice via AI based learning algorithms (Apple, Digitally Narrated Books, Overdrive, etc.)--but even they have issues. This seems to work better with non-fiction or report formats than fiction. This trend is something of an unsettling paradigm shift for traditional voice-actor artists and commercial voice studios (Findaway Voices, etc.).  

[SwampNut]

1 minute ago, ironmike said:

Curiously, the current trend is text to voice via AI based learning algorithms (Apple, Digitally Narrated Books, Overdrive, etc.)--but even they have issues. This seems to work better with non-fiction or report formats than fiction. This trend is something of an unsettling paradigm shift for traditional voice-actor artists and commercial voice studios (Findaway Voices, etc.).  

 

We are less than five years away from pretty decent-quality books on tape being available for nearly every book via AI.

 

[ironmike]

You're not wrong.

 

I'm currently involved in a beta test with two of my fiction books released in text to voice (T2V) format via Digitally Narrated Books (DNA) through Apple and Overdrive. Both are AI female voiced.

One of the things I find may be an issue is dialog attribution in a given scene. For example, character A says something to characters B & C. In text the quoted statement is typically attributed to A by ". . . ," said A. Any reaction or dialog response on the part of B or C requires attribution tags to keep the reader/listener from being confused as to who says what. The narrating AI voice speaks the tags as written . . . somehow that isn't working as well as I'd hoped. I'm not certain the AI voice conveys or evokes the emotions I intended for the involved characters and/or the audience. It is a work in progress, so we'll see.  

[SwampNut]

Ah yeah, clearly you're very familiar, and those are indeed a couple of the challenges for AI.  Because we need it to have it be emotionally intelligent, not just IQ intelligent.

 

[rockmeupto125]

I no longer use it, but I see the results of Dragon text to speech every day at work, and it's bad, possibly even worse than it used to be.  I don't know if people are just getting used to the idea of speaking their notes and now talk conversationally instead of enunciating, or if there are just way more providers who lack proficiency in any form of English.  Some notes are just gibberish but Dragon makes them into words instead of "ggghytytz."

[DaveK]

We use Dragon at work... have been for about 10 years.  Works seamlessly for us.