diff --git a/README.md b/README.md
index 2bb5dca..43549dc 100644
--- a/README.md
+++ b/README.md
@@ -3,39 +3,47 @@
 This is a restructured and rewritten version of [bshall/UniversalVocoding](https://github.com/bshall/UniversalVocoding).
 The main difference here is that the model is turned into a [TorchScript](https://pytorch.org/docs/stable/jit.html) module during training and can be loaded for inferencing anywhere without Python dependencies.
 
-### Preprocess training data
+## Generate waveforms using pretrained models
 
-Multiple directories containing audio files can be processed at the same time.
-
-```bash
-python preprocess.py VCTK-Corpus LibriTTS/train-clean-100 preprocessed
-```
-
-### Train from scratch
-
-```bash
-python train.py preprocessed
-```
-
-### Generate waveforms
-
-You can load a trained model anywhere and generate multiple waveforms parallelly.
+Since the pretrained models were turned to TorchScript, you can load a trained model anywhere.
+Also you can generate multiple waveforms parallelly, e.g.
 
 ```python
 import torch
 
 vocoder = torch.jit.load("vocoder.pt")
+
 mels = [
     torch.randn(100, 80),
     torch.randn(200, 80),
     torch.randn(300, 80),
-]
+] # (length, mel_dim)
+
 with torch.no_grad():
     wavs = vocoder.generate(mels)
 ```
 
-Emperically, if you're using the default architecture, you can generate 100 samples at the same time on an Nvidia GTX 1080 Ti.
+Emperically, if you're using the default architecture, you can generate 30 samples at the same time on an GTX 1080 Ti.
+
+## Train from scratch
+
+Multiple directories containing audio files can be processed at the same time, e.g.
+
+```bash
+python preprocess.py \
+    VCTK-Corpus \
+    LibriTTS/train-clean-100 \
+    preprocessed # the output directory of preprocessed data
+```
+
+And train the model with the preprocessed data, e.g.
+
+```bash
+python train.py preprocessed
+```
+
+With the default settings, it would take around 12 hr to train to 100K steps on an RTX 2080 Ti.
 
-### References
+## References
 
 - [Towards achieving robust universal neural vocoding](https://arxiv.org/abs/1811.06292)