Any good text-to-speech voices besides NeoSpeech?

Hey all — I work at an AI research lab in Seattle where we've built a text-to-speech engine that produces speech that's nearly indistinguishable from the human voice. The voices are far less robotic than Amazon Polly and other providers (we think!) and could be used in production.

​Moderator: spam

I have ample experience with various TTS providers for the development of news reading apps. One advise; take utmost caution when dealing with Neospeech. I have lost thousands of dollars on this TTS platform. When it comes to their VoiceText TTS or SDK , once you have invested in installing it into your environment and are ready to deploy to production they will unilaterally change contract terms. The deductible developer license is actually not deductible from the production license fee and they will add steep royalty costs to their regular pricing. They pulled the plug on my app without notification. I lost $4000 on a developer license and setup apart from the SDK programming costs. Months of correspondence including their CEO Devin Lee but I was never even dignified with an answer. On the bright side; I’m currently quite enthausiastc about AWS Polly. Very transparent and competitive pricing and obviously a much more reliable provider. If you want to know more about my TTS experience with Neospeech Inc. or my current development with AWS Polly drop me a line.

In my opinion, Neospeech voices are high quality but they require you to alter their VTML code in order to get the "perfect" voice you want to produce. Luckily their VT editor has a bunch of options to easily change the voice speed, pitch, break, and volume. I might try that before considering other options.

Other voice options were discussed:

• German Voice Pack (SAPI) for use at Captivate 6: https://forums.adobe.com/message/5179479#5179479#5179479
• High Quality Text to Speech Voices: https://forums.adobe.com/thread/1699495

In the most detailed reply, John-Pierre Cornelissen offered:

Not sure if you already had an answer, but I just did some research for Germen voices too and in the end I found that onlyhttp://www.cereproc.com could provide voices that work with Adobe Captrivate 6.

• http://www.ivona.com: I installed the demo and Captivate 6 does see the voices and allows you to assign them to the TTS sentences. However, when it comes to generating the audio, Captivate picks up the voice that is selected in the control panel » speech section. That's fine if you need a single voice in your project, but if you want a dialog between two or more persons then you have a problem. I have sent an email to query their support for a solution but did not receive a reply.
• http://www.neospeech.com: I have seen topics here in the forum where people say it works. Unfortunately they have very few voices and German is not one of them.
• http://www.loquendo.com: I believe this is the provider of the English voices that come with Captivate 6. I searched their website for a demo or a store to purchase more voices and I couldn't find it. I only found a few pages that say they are now part of Nuance with a broken link to their TTS solutions.
• http://www.nuance.com: like Loquendo, I can't find any page with a TTS demo or a store to purchase more voices. I have sent them an email weeks ago and never received a reply.
• http://www.acapela-group.com: you can't buy from them directly, their website sends you to third party resellers. I visited the website of 2 resellers in my country (Netherlands) but again, I couldn't find a page with a demo or store to purchase.

So, the only provider left for me is http://www.cereproc.com. On their website you can quicly find, purchase and download a voice. All you need to do then is install the voice, enter the license and (re-)start Captivate.

As John-Pierre said, "Hope this helps"

Lizroy, I have been facing similar problems. For me, the comments where robotic, too fast, monotonous and few more which I dare not mention

Frankly I dont have an answer to your question. I haven't found any substitute yet. But I have a work around which is neither easy nor quick.

VTML is VoiceText Markup Language. You can include VTML tags in your TTS (Text to Speech) phrases. For example:

<vtml_speed value="90"> Hello<vtml_break level="2"/> lizroy!</vtml_speed>

The phrase "Hello lizroy" will be voiced at 90% of the normal speed and there is be a bigger pause between "Hello" and "lizroy". You can go to the link below to look at all the possible tags. I generally include varying pauses to make the audio sound more human.

https://ondemand.neospeech.com/vt_eng-Engine-VTML-v3.9.0-3.pdf

As I said, not a easy workaround but works until someone tells us both of a different TTS.