/Audio-effect-replicator-pytorch

replicate audio effects by LSTM. based on [coz-a/Audio-Effect-Replicator](https://github.com/coz-a/Audio-Effect-Replicator).

Primary LanguagePythonMIT LicenseMIT

Audio-effect-replicator-pytorch

replicate audio effects by LSTM. based on coz-a/Audio-Effect-Replicator.

what is it

Train audio effects by setting dry sound as X, effected sound as Y and predict audio effect for new given sound file.

Contents

train.py -- training main program
config.yml -- define parameters and the list of sound file.
predict.py -- predict by using trained model and out .nnp file.
fx_replicator.py -- helper program for train.py and predict.py

Environment

sudo pip3 install -r requirements.txt

Usage

training
python3 train.py

You can use the tensorboard to monitor your learning status.

inferencing using training model
python3 predict.py -i ./data/testdata1-2.wav -m checkpoint\date and time\best_result.pth

you need to prepare WAV file as monoural 32bit signed int(uncompessed)format.

Exercise

Using the created training model,
Inference operation can be performed not only on a PC with high processing power
but also on Jetson Nano and Raspberry pi.