diff --git a/modelscope/pipeline_inputs.py b/modelscope/pipeline_inputs.py
index a10db39e..6d4f7794 100644
--- a/modelscope/pipeline_inputs.py
+++ b/modelscope/pipeline_inputs.py
@@ -1,6 +1,5 @@
 # Copyright (c) Alibaba, Inc. and its affiliates.
 
-import cv2
 import numpy as np
 from PIL import Image
 
@@ -29,7 +28,7 @@ INPUT_TYPE = {
     InputType.IMAGE: (str, np.ndarray, Image.Image),
     InputType.TEXT: str,
     InputType.AUDIO: (str, bytes, np.ndarray),
-    InputType.VIDEO: (str, np.ndarray, cv2.VideoCapture),
+    InputType.VIDEO: (str, np.ndarray, 'cv2.VideoCapture'),
     InputType.BOX: (list, np.ndarray),
     InputType.DICT: (dict, type(None)),
     InputType.LIST: (list, type(None)),
@@ -39,8 +38,13 @@ INPUT_TYPE = {
 
 def check_input_type(input_type, input):
     expected_type = INPUT_TYPE[input_type]
-    assert isinstance(input, expected_type), \
-        f'invalid input type for {input_type}, expected {expected_type} but got {type(input)}\n {input}'
+    if expected_type == 'cv2.VideoCapture':
+        # special type checking using class name, to avoid introduction of opencv dependency into fundamental framework.
+        assert type(input).__name__ == 'VideoCapture',\
+            f'invalid input type for {input_type}, expected cv2.VideoCapture but got {type(input)}\n {input}'
+    else:
+        assert isinstance(input, expected_type), \
+            f'invalid input type for {input_type}, expected {expected_type} but got {type(input)}\n {input}'
 
 
 TASK_INPUTS = {
diff --git a/modelscope/utils/error.py b/modelscope/utils/error.py
index 8128f7b0..bb29aaa0 100644
--- a/modelscope/utils/error.py
+++ b/modelscope/utils/error.py
@@ -125,3 +125,12 @@ You can install it with pip on linux or mac:
 Or you can checkout the instructions on the
 installation page: https://github.com/facebookresearch/fastText and follow the ones that match your environment.
 """
+
+# docstyle-ignore
+EASYNLP_IMPORT_ERROR = """
+{0} requires the easynlp library but it was not found in your environment.
+You can install it with pip on linux or mac:
+`pip install pai-easynlp -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html`
+Or you can checkout the instructions on the
+installation page: https://github.com/alibaba/EasyNLP and follow the ones that match your environment.
+"""
diff --git a/modelscope/utils/import_utils.py b/modelscope/utils/import_utils.py
index 1678776e..b251d107 100644
--- a/modelscope/utils/import_utils.py
+++ b/modelscope/utils/import_utils.py
@@ -288,6 +288,7 @@ REQUIREMENTS_MAAPING = OrderedDict([
     ('scipy', (is_scipy_available, SCIPY_IMPORT_ERROR)),
     ('cv2', (is_opencv_available, OPENCV_IMPORT_ERROR)),
     ('PIL', (is_pillow_available, PILLOW_IMPORT_ERROR)),
+    ('pai-easynlp', (is_package_available('easynlp'), EASYNLP_IMPORT_ERROR)),
     ('espnet2', (is_espnet_available,
                  GENERAL_IMPORT_ERROR.replace('REQ', 'espnet'))),
     ('espnet', (is_espnet_available,
diff --git a/requirements/cv.txt b/requirements/cv.txt
index 338218b0..43eba7f9 100644
--- a/requirements/cv.txt
+++ b/requirements/cv.txt
@@ -21,6 +21,7 @@ moviepy>=1.0.3
 networkx>=2.5
 numba
 onnxruntime>=1.10
+opencv-python
 pai-easycv>=0.6.3.9
 pandas
 psutil
diff --git a/requirements/framework.txt b/requirements/framework.txt
index a86c0cc5..abc08cf1 100644
--- a/requirements/framework.txt
+++ b/requirements/framework.txt
@@ -8,7 +8,6 @@ filelock>=3.3.0
 gast>=0.2.2
 jsonplus
 numpy
-opencv-python
 oss2
 Pillow>=6.2.0
 # for pyarrow 9.0.0 event_loop core dump
diff --git a/requirements/nlp.txt b/requirements/nlp.txt
index 433f70f7..694fc7db 100644
--- a/requirements/nlp.txt
+++ b/requirements/nlp.txt
@@ -5,7 +5,6 @@ ftfy
 jieba>=0.42.1
 matplotlib
 nltk
-pai-easynlp
 pandas
 # protobuf version beyond 3.20.0 is not compatible with TensorFlow 1.x, therefore is discouraged.
 protobuf>=3.19.0,<3.21.0