diff --git a/README.md b/README.md
index 826023f..6991e54 100644
--- a/README.md
+++ b/README.md
@@ -39,18 +39,24 @@ Download [model weights](https://huggingface.co/spaces/Audio-AGI/AudioSep/tree/m
   from pipeline import build_audiosep, inference
   import torch
 
+  # Check if a CUDA-compatible GPU is available, and set the device accordingly
   device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
+  # Build an audio separation model by providing a configuration YAML file, a checkpoint path, 
+  and the chosen device
   model = build_audiosep(
         config_yaml='config/audiosep_base.yaml', 
         checkpoint_path='checkpoint/audiosep_base_4M_steps.ckpt', 
         device=device)
 
+  # Define the paths and information for audio separation
   audio_file = 'path_to_audio_file'
   text = 'textual_description'
   output_file='separated_audio.wav'
 
-  # AudioSep processes the audio at 32 kHz sampling rate  
+ # Note: The 'inference' function processes audio at a 32 kHz sampling rate
+ # Perform audio separation using the 'inference' function with the specified model, audio 
+   file, text, output file, and device 
   inference(model, audio_file, text, output_file, device)
   ```
 
@@ -63,17 +69,25 @@ To load directly from Hugging Face, you can do the following:
   from utils import get_ss_model
   import torch
 
+  # Check if a CUDA-compatible GPU is available and set the device accordingly
   device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
+  # Get the source separation model configuration from the specified YAML file
   ss_model = get_ss_model('config/audiosep_base.yaml')
 
+  # Create an instance of the AudioSep model from a pretrained model checkpoint,
+  # while providing the source separation model and setting the device
   model = AudioSep.from_pretrained("nielsr/audiosep-demo", ss_model=ss_model)
 
+  # Define the paths and information for audio separation
   audio_file = 'path_to_audio_file'
   text = 'textual_description'
   output_file='separated_audio.wav'
 
-  # AudioSep processes the audio at 32 kHz sampling rate  
+
+ # Note: The 'inference' function processes audio at a 32 kHz sampling rate
+ # Perform audio separation using the 'inference' function with the specified model, audio 
+   file, text, output file, and device
   inference(model, audio_file, text, output_file, device)
   ```
 <hr>