Text to Speech Quality Help?

[ToddSmyth]

I am using TTS with Julie, which I think works great. I've learned how to tune problem areas with a combination of the tag language, spaces and punctuation. I think it works fine and so does some of my client team but some are hard to please. Are there any recent developments that might help in terms of TTS quality? Software updates, New voices? New tips?

I notice on the NeoSpeech.com site they play a very subtle piano background audio track which seems to mellow out the TTS a bit? I know Captivate allows a background track and I'm not fond of the stock loops. Do you know what the particular piano loop NeoSpeech is using or where to find it? I've already tried Melody Loops, Shutterstock and Getty.  Thank you

[chuck_jones_1]

I'm not sure about the piano loop, but you may also find some other voices (which I think sound terrific . . . especially the British ones) at http://www.text2speech.org/  Using this will be a bit more cumbersome in that you will need to type your text on that site, produce the mp3 file and then download it - to have to import it into Captivate - so give it a whirl and see if the difference in voice quality is worth it.

Personally, I only use TTS for scratch audio for initial client review - and then use a real human voice for the final product. This way I don't have to wait on changes from a V/O talent, and all of the audio sounds uniform.

I hope this is helpful.

CHUCK

[ToddSmyth]

Thank you. I only see one Male Scottish/British voice and the ones in Captivate seem significantly better than these?

[MaryPetriWest]

I share your interest in better-quality speech agents, and I am scouring this forum for answers... One thing I've learned to do in addition to using the vtml tags (not all of which seem to work, BTW), is to spell words phonetically in order to get the proper pronunciation. For example, Kate says the word "antibodies" strangely; so I spell it "antebaudies." I also make liberal use of  vtml "pause" tags and commas, to break up long, robotic-sounding sentences into more natural-sounding phrases.

Oh, and about "piano background audio track" you mentioned, I don't know but I suspect it's the aural equivalent of a watermark rather than something used to enhance the audio. I could be wrong...

[Paul Wilson CTDP]

This is something that I've got better at over the years, in large part due to the documentation that Neospeech provides. I've found the main thing is to simulate the breaths that a real person would take when speaking. Also, another trick I use is to write as people speak and not how we write. This can be challenging when working with a corporate client who is used to technical, formal and grammatically perfect writing. Humans don't talk that way. One final note is to never have the speech agent refer to themselves. For example, you would want to say "As an employee, it's your job to make safety a part of your day." instead of "As employees, we should always make safety part of our day."

As soon as you hear that computerized voice refer to itself or include itself within the group the illusion is stripped away and your audience is instantly reminded that they aren't listening to a real person.

The following video is a tutorial I did a few years ago which might help some of you out.

Adobe Captivate - Text to Speech Hints and Tips

My newer videos are better quality but not much has changed with how you use it. The Notes panel has changed but the functionality is much the same.

[Paul Wilson CTDP]

By the way, if anyone needs the documentation on Neospeech you can download it from my website using the following link.

https://www.paulwilsonlearning.com/vtml