diff --git a/README.md b/README.md
index 1a1ae63..88f2a39 100644
--- a/README.md
+++ b/README.md
@@ -6,72 +6,73 @@
 ## Desription
 
 The audio embedding operator converts an input audio into a dense vector which can be used to represent the audio clip's semantics.
-This operator is built on top of the VGGish model with Pytorch.
-It is originally implemented in [Tensorflow](https://github.com/tensorflow/models/tree/master/research/audioset/vggish).
-The model is pre-trained with a large scale of audio dataset [AudioSet](https://research.google.com/audioset).
+This operator is built on top of [VGGish](https://github.com/tensorflow/models/tree/master/research/audioset/vggish) with Pytorch.
+The model is a [VGG](https://arxiv.org/abs/1409.1556) variant pre-trained with a large scale of audio dataset [AudioSet](https://research.google.com/audioset).
 As suggested, it is suitable to extract features at high level or warm up a larger model.
 
-```python
-import numpy as np
-from towhee import ops
+## Code Example
 
-audio_encoder = ops.audio_embedding.vggish()
+Generate embeddings for the audio "test.wav". 
 
-# Path or url as input
-audio_embedding = audio_encoder("/audio/path/or/url/")
+ *Write the pipeline in simplified style*:
 
-# Audio data as input
-audio_data = np.zeros((2, 441344))
-sample_rate = 44100
-audio_embedding = audio_encoder(audio_data, sample_rate)
-```
+```python
+from towhee import dc
 
-## Factory Constructor
+dc.glob('test.wav')
+  .audio_decode()
+  .time_window(range=30)
+  .audio_embedding.vggish()
+  .show()
+```
 
-Create the operator via the following factory method
+*Write a same pipeline with explicit inputs/outputs name specifications:*
 
-***ops.audio_embedding.vggish()***
+```python
+from towhee import dc
 
+dc.glob['path']('test.wav')
+  .audio_decode['path', 'audio']()
+  .time_window['audio', 'frames'](range=30)
+  .audio_embedding.vggish['frames', 'vecs']()
+  .select('vecs')
+  .show()
+```
 
-## Interface
+## Factory Constructor
 
-An audio embedding operator generates vectors in numpy.ndarray given an audio file path.
+Create the operator via the following factory method
 
+***audio_embedding.vggish(weights_path=None, framework="pytorch")***
 
 **Parameters:**
 
-​	None.
+​   *weights_path: str*
 
+​   The path to model weights. If None, it will load default model weights.
 
-**Returns**: *numpy.ndarray*
+​   *framework: str*
 
-​	Audio embeddings in shape (num_clips, 128).
+​   The framework of model implementation.
+Default value is "pytorch" since the model is implemented in Pytorch.
 
+## Interface
 
+An audio embedding operator generates vectors in numpy.ndarray given an audio file path or a [towhee audio](link/to/AudioFrame/api/doc).
 
-## Code Example
 
-Generate embeddings for the audio "test.wav". 
+**Parameters:**
 
- *Write the pipeline in simplified style*:
+​   *Union[str, towhee.types.Audio]*
 
-```python
-from towhee import dc
+​   The audio path or link in string.
+Or audio input data in towhee audio frames.
 
-dc.glob('test.wav')
-  .audio_embedding.vggish()
-  .show()
-```
 
-*Write a same pipeline with explicit inputs/outputs name specifications:*
+**Returns**:
 
-```python
-from towhee import dc
+​   *numpy.ndarray*
 
-dc.glob['path']('test.wav')
-  .audio_embedding.vggish['path', 'vecs']()
-  .select('vecs')
-  .show()
-```
+​   Audio embeddings in shape (num_clips, 128).
 
 
diff --git a/vggish.py b/vggish.py
index 69c3e3f..4a91cc5 100644
--- a/vggish.py
+++ b/vggish.py
@@ -27,8 +27,7 @@ from towhee.operator.base import NNOperator
 from towhee.models.vggish.torch_vggish import VGG
 from towhee import register
 
-sys.path.append(str(Path(__file__).parent))
-import vggish_input
+import .vggish_input
 
 warnings.filterwarnings('ignore')
 log = logging.getLogger()
diff --git a/vggish_input.py b/vggish_input.py
index 56e6176..1bc2a13 100644
--- a/vggish_input.py
+++ b/vggish_input.py
@@ -21,8 +21,8 @@ import torch
 import numpy as np
 import resampy
 
-import mel_features
-import vggish_params
+import .mel_features
+import .vggish_params
 
 import torchaudio