Hacker News new | past | comments | ask | show | jobs | submit login

> festival is currently hardcoded as the text-to-speech engine.

How is that, quality wise? I tried to get festival working once, but failed.

Also tried espeak, but found it to be not very good.

Finally settled on pico2wave with the "-l=en-GB" flag to make it use the British Lady voice, which seemed really really excellent compared to everything else I was able to try. Edit: I excluded online TTS because for what I was doing it needed to be local TTS.

> "-l=en-GB" flag

This actually changes everything! With the default US voice, it's far from the same quality level.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact