Dear Text to Speech and open source experts:
For a toy / hobby project (non-commercial), I’m trying to find a suitable Text to Speech system for Dutch that I can run on GNU/Linux. So far, the situation does not look very promising. I’ve tried eSpeak, but using it for Dutch is not as good as I expect. I made my experiment using a file “computer.txt” that has the following contents:
Een computer is een apparaat waarmee gegevens volgens formele procedures zoals algoritmen kunnen worden verwerkt. Meestal wordt met het woord computer een elektronisch, digitaal apparaat bedoeld, maar er bestaan ook mechanische en analoge computers.
$ espeak -vnl+7 -s 170 -f computer.txt
The resulting voice output:
My eSpeak version is:
$ espeak --version eSpeak text-to-speech: 1.48.03 04.Mar.14 Data at: /usr/lib/x86_64-linux-gnu/espeak-data
I’ve tried tweaking various parameters such as speed, pitch, voice but could not get something that is much better.
The advantages of eSpeak: It doesn’t need any Internet connection when running. The disadvantage, as can be heard above is low quality for Dutch.
When I say eSpeak has low quality for Dutch, I compare it to Google Translate’s output. You can try it for yourself at https://goo.gl/8urVXb (by clicking on the loudspeaker icon on the bottom left). Or you can listen my recording of it below:
The advantages of Google Translate’s Text to Speech system: Higher quality for Dutch, but the disadvantage is that it requires an Internet connection and the API is probably paid.
Maybe there’s something better, some solution that I can install and run on GNU/Linux, control / drive programmatically (e.g. via Python, Bash, etc.) and does not charge me a lot of money (because I’m not building a commercial product). Any ideas?
Notes: For the curious reader, to record the audio from espeak and Google Translate using Audacity, I used Pulse Audio Volume control. Following resources helped me to set it up: