Larynx is the text-to-speech engine for the Rhasspy project. It runs as a service, so you can send text and get a .wav file back in return.
FYI: last time I tried this, it was broken and there were no voices available. If you know how to fix it, please send a PR.
Run make config to configure the larynx domain name and
username/password.
See AUTH.md for information on adding external authentication on top of your app.
Run make install to deploy the app.
Run make open to open the page in your browser.
Larynx serves a web browser test page, but you can also use the service API directly:
curl https://larynx.example.com/api/tts \
-u larynx:password \
-G \
--data-urlencode 'text=this is a test' \
--data-urlencode 'voice=en-us/mary_ann-glow_tts' \
--data-urlencode 'vocoder=hifi_gan/universal_large' \
--data-urlencode 'denoiserStrength=0.002' \
--data-urlencode 'noiseScale=0.333' \
--data-urlencode 'lengthScale=1' \
--data-urlencode 'textLanguage=' \
--data-urlencode 'inlinePronunciations=true' \
--compressed | aplay -r 22050 -c 1 -f S16_LE