HACKER Q&A
📣 ndehouche

HN: What is currently the most realistic text-to-speech service?


I have this idea of making short videos defining one concept. I have used Google's https://youtu.be/Lt6bwkR7bUI Also tried Amazon Polly but it seems worse. Are there currently better alternatives realism-wise?


  👤 uberman Accepted Answer ✓
Is this the google (wavenet) version you tried?

https://cloud.google.com/text-to-speech

MS offers a similar service:

https://azure.microsoft.com/en-us/services/cognitive-service...

Both default to voices seem a little tinny to me, but both are customizable. We use one of these services for a US audience with a European accent as we find that further obfuscates the generated speach.


👤 just-juan-post
What exactly do you mean by "realism-wise"?

Elaborate so that we know what you want. Perhaps some products can be tuned to your liking.


👤 taf2
Amazon Polly is pretty nice