@GregorR what will be changes for training the model and inference code for support 16khz input
audio instead of 48khz?

I have two questions:

1. Could you friends be kind enough to share a sample ffmpeg execution script for these rnn models? I can later add the sample script in to the readme file so everyone can see.
2. I will be using this on mobile devices. Could you please suggest me if this is feasible? How is the performance generally compared to non neural denoicing algorithms?

Hi, I'm studying about RNN, I want to know how what are performance comparison values(accuracy) for these models? Thank you.

I am using speech reorganization to generate srt files for large video files (in various languages...with lot of background voices, music, chit-chat, ...blah blah....just what is present in the most of the films)....kindly suggest which model to use to filter all except speeches.

I cannot make it out from your table in read me .... is it somnolent-hogwash? or I need others also to use and or in specific order with ffmpeg?
Thanks.

Hi, can you talk? How to create the model .rnnn step by step? i have weights.hdf5 file and i don't know how to take a model. Can you write please...

Is it possible to convert the original model to something that ffmpeg's arnndn can use?

Hi
I am trying to use somnolent-hogwash-2018-09-01 model for rnnoise_nu_demo

running the following examples/rnnoise_nu_demo 1 1 ./rnnoise-models/somnolent-hogwash-2018-09-01/sh.rnnn ../rnnoise/file_in.raw file_out.raw

But getting - Model not found!