4 ms·
Ask HN: 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 6y agoIs this the google (wavenet) version you tried? https://cloud.google.com/text-to-speech https://cloud.google.com/text-to-speech MS offers a similar service: https://azure.microsoft.com/en-us/services/cognitive-services/text-to-speech/ 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 6y agoWhat exactly do you mean by "realism-wise"? Elaborate so that we know what you want. Perhaps some products can be tuned to your liking.
- ndehouche 6y agoMy problem with Google is that sometimes the voice becomes "metallic" on some portions of the text. Not having that would be nice. Also perhaps some randomness. When saying the same few words twice, there could be a little variety. More generally, I am also wondering who has the best tech currently (assuming the answer is settled enough).
- harpreetsahota 6y agoFor my podcast transcripts I use Sonix.ai. I don't know if this is exactly what you're looking for, but it's worked great for me. I upload the file, it does the transcription with some pretty decent accuracy. It's 22/month and 5/hour of transcription.
- taf2 6y agoAmazon Polly is pretty nice