Each time I make rhasspy speaking, being with api or interface test, I ear a scratch (sort ‘pttr’) after the sentence.
I never heard this with snips tts and use same respeaker2mic pi hat with seed drivers. The only difference is pi3->pi4 stretch->buster.
This doesn’t happend with:
pico2wave -l fr-FR -w test.wav "Petit test"
Doesn’t happend either when playing Wake WAV and Recorded WAV (which I changed for snips ones )
Anyone also hearing this ? Any idea ?
Exactly same problem (but on a Pi3, Buster). Was opening an issue on GitHub, then checked for the same problem here. I also tested with aplay and pico2wave and the crack isn’t there. No idea on what could be the cause.
Any help appreciated. Also any test I can do to help I’m available.
P.S. Many many many thanks to the developer of this big and awesome project. Found this after Snips announcement and wish I used it from the first time.
@synesthesiam any idea what could cause this ?
Yeah, I went ahead and pushed a fix to master. I hadn’t noticed the pop before; I was trying to avoid writing a temporary WAV file (which
pico2wave just insists on doing) with a “clever” hack.
So the hack is now removed
As an alternative, NanoTTS might help as it can output raw audio data directly to stdout (it is raw audio data though).
I’d like to replace PicoTTS with this in the future.