Audio Conversion through AI

[BobbyBrandt]

A few questions. Maybe some of these need to be answered by Laurel, but I'll throw them out here to start:

1. The number and quality of artificial intelligence platforms for converting text to speech are increasing every day. (Amazon "Polly" is one that I have played around with recently) Does anyone know if these AI-generated audio stories have been posted on Litertotica and what the feedback was?
2. How do the admins here feel about an audio story that is derived entirely from a story by the same author previously published on Literotica?
3. Since "Audio Sex Stories" is a category of its own, what is the best way to categorize an audio submission to attract readers interested in that genre? Tags?

 

[NotWise]

I'll take a shot at this, though I haven't posted an audio of my own.

A few questions. Maybe some of these need to be answered by Laurel, but I'll throw them out here to start:

1. The number and quality of artificial intelligence platforms for converting text to speech are increasing every day. (Amazon "Polly" is one that I have played around with recently) Does anyone know if these AI-generated audio stories have been posted on Litertotica and what the feedback was?

I don't know. I've used one (Google Translate) to convert my WIP (about 18K words at the time) and I would never use something like that to present my work. There are just too many problems. I don't know how it compares to other applications.

I did think it was an interesting excercise and a good way to catch a few kinds of errors, like phrasing. If the AI is confused by it, then the human reader probably will be too.

2. How do the admins here feel about an audio story that is derived entirely from a story by the same author previously published on Literotica?

JuanaSalsa's "Hog Roast" is one of those, posted recently. I haven't heard of any kickback from Laurel, and I doubt there would be.

Keep in mind, though, that the erotic audio on Lit is usually pretty different from a story read aloud. There are exception (like JuanaSalsa's), but most are short--often around twenty minutes--and consist of one character talking to the listener. They're different from audio books, which might be what you have in mind.

3. Since "Audio Sex Stories" is a category of its own, what is the best way to categorize an audio submission to attract readers interested in that genre? Tags?

They seem to use title, short description, and tags. Often in the title or short description you'll see something like [M4F] or [F4M], meaning male actor for a female listener and female actor for a male listener, respectively.

 

[BobbyBrandt]

I decided to give it a try by converting one of my published stories (The Voice in my Head) to an audio file. I chose this story because it is one of my shorter ones at just over 5,000 words and generated a 25-minute audio file of approximately 12 megabytes.

I used Amazon Polly for the voices and then used Audacity to merge the multiple sound files. It isn't broadcast quality, but it turned out better than I expected. The AI-generated voice for the main character is especially good. I need to play around with SSML tags in Polly some more and I think that will do the trick in the future.

When it came to submitting the audio file, I did have to rename it due to the previously published story having the same name. I just add "(Audio)" to the title and it was accepted, though still pending.

Now we wait for the reception.

 

[NotWise]

I decided to give it a try by converting one of my published stories (The Voice in my Head) to an audio file. I chose this story because it is one of my shorter ones at just over 5,000 words and generated a 25-minute audio file of approximately 12 megabytes.

I used Amazon Polly for the voices and then used Audacity to merge the multiple sound files. It isn't broadcast quality, but it turned out better than I expected. The AI-generated voice for the main character is especially good. I need to play around with SSML tags in Polly some more and I think that will do the trick in the future.

When it came to submitting the audio file, I did have to rename it due to the previously published story having the same name. I just add "(Audio)" to the title and it was accepted, though still pending.

Now we wait for the reception.

Let us know when it goes live. It'll be interesting.

Two-hundred words a minute seems like pretty fast talking. Is that a setting?

 

[BobbyBrandt]

Let us know when it goes live. It'll be interesting.

Two-hundred words a minute seems like pretty fast talking. Is that a setting?

The rate I used was the default, and it sounds like a pretty normal conversational rate. There are SSML tags that adjust the rate faster or slower, but I didn't use any for this first attempt.

 

[BobbyBrandt]

Let us know when it goes live. It'll be interesting.

Two-hundred words a minute seems like pretty fast talking. Is that a setting?

Here is the link

 

[ReadWriteDelete]

It's true text-to-speech systems are getting better, but noone in the voice-over industry is worried. Even with hours of tweaking using markup to add inflection, pauses and emphasis the results have jarring robotic characteristics.

At Librivox, artificial voices are specifically not allowed for their audiobooks. Same for commercial audiobooks like those from Audible.

Many Youtube videos use of text-to-speech when the creator doesn't speak English well or needs to hide their voice. Many of those seem to be done using NaturalReader. It's good enough for that purpose, but obviously artificial and to most ears kind of annoying.

I think even an untrained human reader beats the best TTS every time, especially for erotica where, just like with sex itself, the performance makes all the difference

 

[NotWise]

That didn't take long to post.

Polly gave you a better product than I expected, but it still has a lot of problems -- especially with phrasing. The main character's voice was impressively good at first, but to me it became mechanical as the story went on. All of the other voices had significant problems.

I did feel like 200 words/minute was a little too fast.

 

[InMyHandsAudio]

WOw! This it really interesting. Thank you for all the info on this subject.
I am very curious to listen to the results.

 

[BobbyBrandt]

It's true text-to-speech systems are getting better, but noone in the voice-over industry is worried. Even with hours of tweaking using markup to add inflection, pauses and emphasis the results have jarring robotic characteristics.

At Librivox, artificial voices are specifically not allowed for their audiobooks. Same for commercial audiobooks like those from Audible.

Many Youtube videos use of text-to-speech when the creator doesn't speak English well or needs to hide their voice. Many of those seem to be done using NaturalReader. It's good enough for that purpose, but obviously artificial and to most ears kind of annoying.

I think even an untrained human reader beats the best TTS every time, especially for erotica where, just like with sex itself, the performance makes all the difference

As I said, I'm playing around with the technology and published this one as a sample of what could be accomplished. In this particular case, I used multiple AI-generated voices, and as you mentioned the quality of some are far superior to others. I purposely chose a shorter story for this conversion.

I am doing another conversion right now that will use a single voice to read a story that has a lot of dialogue tags. The single-voice "story-reader" approach is more suitable for this type of story, and I can use a voice that is less robotic throughout. The problem is that this story is almost 100k words long and will need to be broken up by chapter in audio format.

 

[shadowdriva]

I use spoken content under the iPhone accessibility to listen to stories. Is there a way to merge all sections of a story into one. Better yet is there a reader that I can import the story into? On Android, I can use the app and share parts of the story with voice aloud reader which allows me to listen to an entire story. Any method of listening to stories would be greatly appreciated.

 

[DoctorSmith8]

A few questions. Maybe some of these need to be answered by Laurel, but I'll throw them out here to start:

1. The number and quality of artificial intelligence platforms for converting text to speech are increasing every day. (Amazon "Polly" is one that I have played around with recently) Does anyone know if these AI-generated audio stories have been posted on Litertotica and what the feedback was?
2. How do the admins here feel about an audio story that is derived entirely from a story by the same author previously published on Literotica?
3. Since "Audio Sex Stories" is a category of its own, what is the best way to categorize an audio submission to attract readers interested in that genre? Tags?

I've been using "text to speech" to voice my stories for 6 years. In fact that's what got me started on my writing journey. All of the software and voices (about 25) are located in my iPad. So everything including the writing is accomplished right there. Once the multi character dialogue work is finished it gets transferred to my PC using a free multitrack recorder called Audacity. Then I'm ready to add sound effects and some mood music. (Non copyright of course) I use a separate voice for each character and write in a screenplay format.

Right now I have 26 of them stored on-line for free listening. My provider counts the amount of download traffic I receive so I can check the popularity of each one. After the total exceeded 10,000 I quit counting. So far I have been unsuccessful loading anything onto Lit.

here is an example...

"The Machine" A college student has his first experience at a sperm bank.

https://audiomack.com/adventures-in-the-stratosphere/song/16019404

 

[BobbyBrandt]

I've been using "text to speech" to voice my stories for 6 years. In fact that's what got me started on my writing journey. All of the software and voices (about 25) are located in my iPad. So everything including the writing is accomplished right there. Once the multi character dialogue work is finished it gets transferred to my PC using a free multitrack recorder called Audacity. Then I'm ready to add sound effects and some mood music. (Non copyright of course) I use a separate voice for each character and write in a screenplay format.

Right now I have 26 of them stored on-line for free listening. My provider counts the amount of download traffic I receive so I can check the popularity of each one. After the total exceeded 10,000 I quit counting. So far I have been unsuccessful loading anything onto Lit.

here is an example...

"The Machine" A college student has his first experience at a sperm bank.

https://audiomack.com/adventures-in-the-stratosphere/song/16019404

Good story but the characters sound robotic and obviously AI-generated.

I have found with Amazon Polly and Audacity, I can create audio books that sound extremely lifelike. I might follow your lead, however and look into uploading them onto audiomac to give easiere access to them.

 

[DoctorSmith8]

Good story but the characters sound robotic and obviously AI-generated.

I have found with Amazon Polly and Audacity, I can create audio books that sound extremely lifelike. I might follow your lead, however and look into uploading them onto audiomac to give easiere access to them.

They were created 6 years ago before AI was a thing. But yeah, that's as good as it got back then. At least I was able to communicate a story. Now I have been sticking to just a regular written format.

 

[BobbyBrandt]

They were created 6 years ago before AI was a thing. But yeah, that's as good as it got back then. At least I was able to communicate a story. Now I have been sticking to just a regular written format.

I wanted to thank you for the tip about Audiomack. It looks like it might be a useful site for me.

Here is a link to a reading from one of my stories that I converted to audio using Amazon Polly, edited through Audacity, and then posted on Audiomack. It's not perfect, and if I decide to convert the entire book, it would take some adjustments, but I think it illustrates the progress made with the AI technology.

 

[DoctorSmith8]

I wanted to thank you for the tip about Audiomack. It looks like it might be a useful site for me.

Here is a link to a reading from one of my stories that I converted to audio using Amazon Polly, edited through Audacity, and then posted on Audiomack. It's not perfect, and if I decide to convert the entire book, it would take some adjustments, but I think it illustrates the progress made with the AI technology.

Yes, that is a great sounding voice. Tech has come a long way! Try this one...

"A Vampire Called Julie" The story of a modern day female vampire and her unique method of feeding.

https://audiomack.com/adventures-in-the-stratosphere/song/a-vampire-called-julie