4 ms·
I was looking at some of the text-to-speech offerings on platforms like aws recently, and this is way better. You can still tell its a robot, but it is more nat
by ksaj 6y ago
I was looking at some of the text-to-speech offerings on platforms like aws recently, and this is way better. You can still tell its a robot, but it is more natural and doesn't sound like it is droning.
- achow 6y agoGoogle Wavenet sounds much more natural than this. https://cloud.google.com/text-to-speech#section-2 https://cloud.google.com/text-to-speech#section-2
- ksaj 6y agoThanks. It's definitely very good, and might work well for my projects. Natural flow is more important for me than timeline (although I can see that as quite important most types of videos), and I was already hoping to use SSML for the granular control. I quite like the fact that the examples use curl, since that closely matches how I'd rather feed it. I'll have to see what its maximum input length is. Most of the texts I'm working with are about 66,000 characters and up at a time (and then more if I use SSML). Unsurprising that Google would have voices of that calibre. I was surprised that Amazon's voices sound so poor. And they weren't the worst that I found on my ways. My Mac's built-in voices are better than those.