diff --git a/project/midi.py b/project/midi.py
index 04ca47f..7e03254 100644
--- a/project/midi.py
+++ b/project/midi.py
@@ -39,6 +39,7 @@ def to_midi(samples, output_path=settings.generated_midi_path, program=0, tempo=
     return_midi = roll.Multitrack(tracks=tracks, tempo=tempo, downbeat=[0, 96, 192, 288], beat_resolution=beat_resolution)
     roll.write(return_midi, settings.generated_midi_path)
 
+# todo: this function is running too slow.
 def delete_empty_samples(sample_pack):
     print('Deleting empty samples...')
     temp_sample_pack = sample_pack
@@ -64,6 +65,7 @@ def main():
                 continue
             sample_pack = np.concatenate((midi_samples, sample_pack), axis=0)
 
+        # I commented out this line, because it was too slow
         # sample_pack = delete_empty_samples(sample_pack)
 
         np.savez_compressed(settings.samples_dir, sample_pack)
diff --git a/project/readme b/project/readme
new file mode 100644
index 0000000..b65f2d1
--- /dev/null
+++ b/project/readme
@@ -0,0 +1,16 @@
+## MUSIC GENERATION USING DEEP LEARNING ##
+## AUTHOR: CEZARY PUKOWNIK
+
+How to use:
+
+1. Use midi.py to export data from midi files
+
+  ./midi.py export <midi_folder_path> <output_path>
+
+2. Use train.py to train a model (this can take a while)
+
+  ./train.py <input_training_data> <model_save_path>
+
+3. Use generate.py to generate music from trained models
+
+  ./generate.py <model_weights_path> <output_path>