Dereverb for speech?
Hi @Sucial ! Thanks so much for all of your work!
Just curious is this model is intended / suitable to be used for removing reverb in speech (i.e. non-music use case)? Or is it intended to only remove reverb in vocals (as in, extracted vocals from a music track)?
Thanks again!
It may not be possible to remove reverb in speech, I trained this model using stereo singing.
So it may only works on extracted vocals from a music track. In addition, I have also tested that the model cannot remove mono reverberation and only works on stereo reverberation.
Got it, okay makes sense. I noticed that for mono (speech) tracks, I needed to apply some 'pseudo-stereo' effects first in order for it to have some effect. Anyway, thanks!
Try this model: https://huggingface.co/anvuew/dereverb_mel_band_roformer/blob/main/dereverb_mel_band_roformer_mono_anvuew_sdr_20.4029.ckpt with config: https://huggingface.co/anvuew/dereverb_mel_band_roformer/blob/main/dereverb_mel_band_roformer_anvuew.yaml
which is trained by anvuew and it can deal with mono tracks.
Ah great, I will try it, thank you!