Text to Speech - Again

Jan 28, 2013

I know there have been several posts about Text to Speech, and I have read most of them.  I am having a hard time finding any of the voices out there that actually sound natural.  I am remaking our compliance training and using Storyline to do so.  I am creating the voices out of Captivate 6, which uses some of the same voices a lot of the different voice engines out there use.  The problem is, they are just so bland and robotic.  No emotion, mispronunciations, automaton sounding.  I dont even know if there is a solution for this, but I thought I would ask one more time. 

Is there any truly natural sounding text to speech out there?  Thanks!

8 Replies
Steve Flowers

No TTS I've found is perfect. They've gotten better over the past 5 years. This is the one I've used in the past:
http://www.ivona.com/us/

Loquendo offers emotes and shifts within their TTS but I haven't found any of the shelf / affordable tools that use the Loquendo engine:

http://www.loquendo.com/en/demo-center/interactive-tts-demo/ 

Bob S

Hi Eric,

I would love to hear some additional ideas from some other folks here as well on the exact same question.

Here is what I found in my research on the topic thus far...

Good - Ivona

Pretty Good - Loquendo

Pretty Good - NeoSpeech

Couple of other things...


  • None of them are perfect and all have some degree of robotic flavor to them. That can be fine in today's world.

  • Seems like the UK female voices often wind up being some of the best (most human) for each company. Interesting.

  • Many of them respond differently depending on how you spell, punctuate, etc your sentences. Don't be discouraged if they don't get it right straight away. Try typing alternatives and you can often find something suitable.

Eager to hear more from some other folks...

Bob

James Brandwood

I would also love to hear what people are currently using.

I've trialled a few of these, hoping to save time on projects but I have never found one that sounds human, particularly when using longer (in syllables) or more technical words like 'simulation'. I do agree with Bob that the female UK voices sound the best.

I would also like to add that I do not sound like the Australian accent option on Ivona. I think if I used that on a training module people would think I just compiled sound bites of Paul Hogan from Crocodile Dundee (80's movie where he put on a really thick fake accent).

Eric Traugott

James Brandwood said:

I would also love to hear what people are currently using.

I've trialled a few of these, hoping to save time on projects but I have never found one that sounds human, particularly when using longer (in syllables) or more technical words like 'simulation'. I do agree with Bob that the female UK voices sound the best.

I would also like to add that I do not sound like the Australian accent option on Ivona. I think if I used that on a training module people would think I just compiled sound bites of Paul Hogan from Crocodile Dundee (80's movie where he put on a really thick fake accent).

Lol!  I am in the process of trying out ispeech www.ispeech.org  So far, they sound quite a bit better than the voices I have been pulling out of Captivate. 
Anna Vilarnau

Hi Eric,

I know we spoke yesterday about our human-quality Text To Speech, but I wanted to take the opportunity to share our demo with the community, so everyone can try our voices: http://www.ispeech.org/text.to.speech

I would also like to highlight that we can adjust and improve the pronunciation, and we offer custom voices on demand. If anyone is interested I can help, you can send me an email to anna [at] ispeech [dot] org

Richard Gordon

Hi Eric, you might want to check out the voices at YAKiToMe! http://www.yakitome.com. There are many free, natural-sounding voices to choose from. The software allows you to convert documents of any length for free, unlike the other so-called "free" websites out there. And there's built-in pronunciation editing software to make it easy for you to correct pronunciations you don't agree with. You can download the resulting audio in standard file formats, like .mp3 and .ogg.

This discussion is closed. You can start a new discussion or contact Articulate Support.