- Zach Wener
Uberduck is a website where anyone can create a text-to-speech voice. And people create a LOT of text-to-speech voices. At the time that I'm writing this, over 3,000 voices are live on Uberduck and all of them were created by our community.
But some of the voices sound better than others—the speech they generate is more natural or more expressive. It turns out that creating a text-to-speech voice requires skill, dedication, attention to detail, and even a bit of luck at times. With over 3,000 voices to choose from, it can be difficult to find the good ones.
That's why we added a voice feedback system. When you generate audio on Uberduck, you can click the thumbs up or thumbs down button to give the voice creator feedback that the output either met your expectations or needs improvements. Make sure you use the feedback system when you use Uberduck—giving feedback helps make Uberduck better for everyone!
You can find the most popular and highest-rated voices in the Voice Leaderboard.
To find the best voices, look for voices that have a high quality score and have been used many times. You can sort the table by Quality and then do a secondary-sort by Usage by first clicking Quality, then holding the
Shift key and clicking Usage.
The usage counts reset after someone updates the voice, so if you improve a voice over time, you don't have to worry about previous low ratings weighing down the score.
If you liked this post and are excited about text to speech and machine learning, hop in the Discord and join us.