diff --git a/README.md b/README.md
index f8f0ff7..4937718 100644
--- a/README.md
+++ b/README.md
@@ -40,7 +40,13 @@ If the environment is installed correctly, you can denoise an audio file by runn
     
 A ".wav" file with the denoised version, as well as the residual noise and the original signal in "mono", will be generated in the same directory as the input file.
 ## Training
-TODO
+To retrain the model, follow the instructions:
+
+Download the [Gramophone Noise Dataset](http://research.spa.aalto.fi/publications/papers/icassp22-denoising/media/datasets/Gramophone_Record_Noise_Dataset.zip), or any other dataset containing recording noises.
+
+Prepare a dataset of clean music (e.g. [MusicNet](https://zenodo.org/record/5120004#.YnN-96IzbmE))
+
+
 ## Remarks
 
 The trained model is specialized in denoising gramophone recordings, such as the ones included in this collection https://archive.org/details/georgeblood. It has shown to be robust to a wide range of different noises, but it may produce some artifacts if you try to inference in something completely different.