4

My company recently got contracted to build a children's website in which the characters speak to the kids throughout the site. In an effort to make the site more personalized and child friendly, we are looking to have the characters speak in "character-like" voices. The problem we're running into is that the pitch, speed, duration, and rate controls most of the text to speech platforms offer don't give enough control to get the voices sounding like storybook characters and not like creepy old men.

We tossed around the idea of pre-recording voiceovers and playing them back as needed but the site owners requested the ability to change the spoken content on the fly. We've also tossed around the idea of saving the text to speech outputs to a file and filtering the audio with something like FFMPEG. That's a lot of work if theres something out there that will do it already programmatically.

My question is if anyone knows of a text to speech engine with custom voice controls or a way to filter the built-in voices.

The site is being built primarily in JavaScript and PHP and run on a Linux server.

1

This guide suggests using Festival, eSpeak or FreeTTS with voices from FestVox, MBROLA, CMU Arctic or Nitech HTS. Alternative software should come with different voices.

1

Amazon Polly has three children's voices in US English only. They are "neural" voices, so they sound very natural.

1

There is a website called https://www.resemble.ai/ and it allows you to make your own voices. However, you will need a child and a couple of hours (a good two days, perhaps). Resemble AI also offers pre-made voices, if any are good enough for you.

You can change the content on the fly using their JS API: https://app.resemble.ai/docs

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.