Conversation
|
Also, is it possible to create the sounds with a little higher volume? The existing sounds are a bit quiet. |
|
Second part of the process is to normalise them all. Can't realistically go any louder than what that will be as the audio will start distorting then, and quite badly in some cases. |
|
Actually, if you mean the current phrases as per the zip file releases (not the ones in the repo - those are pre-normalisation as that is done at release time) then they are already max volume before they will start to distort. I won't be able to change the pitch or speed atm - that requires SSML formatted phrases... which I was initially going to say isn't going to happen since AFAICT that would then lock this voice pack system to Microsoft Azure - whereas the CSV files are pretty TTS agnostic, but I might actually be able to fake that judging form how the formatting is done Azure-Samples/cognitive-services-speech-sdk#742 (comment) - so will investigate that a bit further. |
|
OK, I understand. I'll crank up the volume instead :) |
To provision a request that was in #51
f64aa2a to
bd64ff3
Compare
Some changes to generate better pronunciation.
The translations would be a little easier to hear if both speed and pitch could be set to 0.9.
Don't know where/how to change these parameters but Swedish Sofie Neural with pitch 0.9 and speed 0.9, please.