Improve the quality of the voice

[okulovsky]

The samples that are generated by CoquiTTS are a bit noisy, there is some "sand" in them. We need some solution to enhance the quality, which e.g. https://voice.ai/tools/audio-enhancer does quite well (but it's only available on Windows, so, not compatible on docker)

We need to find a tool that can do it within the Linux container, and implement it in BrainBox.

ResembleEnhance is not a solution as it can only process the English texts

[okulovsky]

Maybe one of these will do
https://github.com/modelscope/ClearerVoice-Studio
https://github.com/speechbrain/speechbrain.git
those are called speech enhancement systems btw

[okulovsky]

Clearer Voice doesn't solve the issue

[okulovsky]

SpeechBrain pretrained models also don't solve the issue