search

MycroftAI / mimic-recording-studio

Mimic Recording Studio is a Docker-based application you can install to record voice samples, which can then be trained into a TTS voice with Mimic2
Apache License 2.0
499 stars 116 forks source link

Could we use a NN to denoise the audio? #50

Open DrAlta opened 3 years ago

DrAlta commented 3 years ago

It says in t doc that any auditable noise will make bade samples.

Could we use something like https://github.com/deezer/spleeter

To remove the background noise?

krisgesling commented 3 years ago

Hey there,

It seems like there's a need for more built-in tools like this.

I'd strongly recommend using them as post-processing tools rather than operating on the input stream or replacing the raw files. If you retain your original recordings you can always process them again using different parameters.

I'm not sure that Spleeter would be my first choice, and would require you to have training data for what you are trying to separate out. However a more generic noise removal tool could be very helpful. I'd assume that most people use more traditional tools like Audacity.

marshalleq commented 1 year ago

The ML noise removal tools are absolutely magic and much better than traditional like audacity in my opinion - these are coming pertained Ultimate Vocal remover being a very good example. Traditional tools won't catch a lot of the noises that we experience when we don't have a professional recording studio in our houses.