diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 586cc83..1981786 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -7,7 +7,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python: [ "3.9", "3.10", "3.11" ]
+        python: [ "3.9", "3.10", "3.11", "3.12" ]
         backend: [ tensorflow, jax ]
         include:
           - backend: tensorflow
diff --git a/README.md b/README.md
index 4e6313f..1d0a76b 100644
--- a/README.md
+++ b/README.md
@@ -61,21 +61,23 @@ For more details, please refer to the [API](#api) section.
 ```python
 import opennsfw2 as n2
 
-# To get the NSFW probability of a single image.
-image_path = "path/to/your/image.jpg"
+# To get the NSFW probability of a single image, provide your image file path,
+# or a `PIL.Image.Image` object.
+image_handle = "path/to/your/image.jpg"
 
-nsfw_probability = n2.predict_image(image_path)
+nsfw_probability = n2.predict_image(image_handle)
 
-# To get the NSFW probabilities of a list of images.
-# This is better than looping with `predict_image` as the model will only be instantiated once
-# and batching is used during inference.
-image_paths = [
+# To get the NSFW probabilities of a list of images, provide a list of file paths,
+# or a list of `PIL.Image.Image` objects.
+# Using this function is better than looping with `predict_image` as the model 
+# will only be instantiated once and batching is done during inference.
+image_handles = [
   "path/to/your/image1.jpg",
   "path/to/your/image2.jpg",
   # ...
 ]
 
-nsfw_probabilities = n2.predict_images(image_paths)
+nsfw_probabilities = n2.predict_images(image_handles)
 ```
 
 ## Video
@@ -154,8 +156,8 @@ Create an instance of the NSFW model, optionally with pre-trained weights from Y
 ### `predict_image`
 End-to-end pipeline function from the input image to the predicted NSFW probability.
 - Parameters:
-  - `image_path` (`str`): Path to the input image file. 
-    The image format must be supported by Pillow.
+  - `image_handle` (`Union[str, PIL.Image.Image]`):
+    Path to the input image file with a format supported by Pillow, or a `PIL.Image.Image` object.
   - `preprocessing`: Same as that in `preprocess_image`.
   - `weights_path`: Same as that in `make_open_nsfw_model`.
   - `grad_cam_path` (`Optional[str]`, default `None`): If not `None`, e.g., `cam.jpg`,
@@ -171,8 +173,9 @@ End-to-end pipeline function from the input image to the predicted NSFW probabil
 ### `predict_images`
 End-to-end pipeline function from the input images to the predicted NSFW probabilities.
 - Parameters:
-  - `image_paths` (`Sequence[str]`): List of paths to the input image files. 
-    The image format must be supported by Pillow.
+  - `image_handles` (`Union[Sequence[str], Sequence[PIL.Image.Image]]`):
+    List of paths to the input image files with formats supported by Pillow,
+    or list of `PIL.Image.Image` objects.
   - `batch_size` (`int`, default `8`): Batch size to be used for model inference. 
     Choose a value that works the best with your device resources.
   - `preprocessing`: Same as that in `preprocess_image`.
diff --git a/opennsfw2/_inference.py b/opennsfw2/_inference.py
index 0d3c187..1e1652d 100644
--- a/opennsfw2/_inference.py
+++ b/opennsfw2/_inference.py
@@ -2,7 +2,7 @@
 Inference utilities.
 """
 from enum import auto, Enum
-from typing import Any, Callable, List, Optional, Sequence, Tuple
+from typing import Any, Callable, List, Optional, Sequence, Tuple, Union
 
 import cv2
 import numpy as np
@@ -28,18 +28,24 @@ def _update_global_model_if_needed(weights_path: Optional[str]) -> None:
         global_model_path = weights_path
 
 
+def _load_pil_image(image_handle: Union[str, Image.Image]) -> Image.Image:
+    if isinstance(image_handle, Image.Image):
+        return image_handle
+    return Image.open(image_handle)
+
+
 def predict_image(
-        image_path: str,
+        image_handle: Union[str, Image.Image],
         preprocessing: Preprocessing = Preprocessing.YAHOO,
         weights_path: Optional[str] = get_default_weights_path(),
         grad_cam_path: Optional[str] = None,
         alpha: float = 0.8
 ) -> float:
     """
-    Pipeline from single image path to predicted NSFW probability.
+    Pipeline from single image handle to predicted NSFW probability.
     Optionally generate and save the Grad-CAM plot.
     """
-    pil_image = Image.open(image_path)
+    pil_image = _load_pil_image(image_handle)
     image = preprocess_image(pil_image, preprocessing)
     _update_global_model_if_needed(weights_path)
     assert global_model is not None
@@ -56,9 +62,9 @@ def predict_image(
     return nsfw_probability
 
 
-def _predict_from_image_paths_in_batches(
+def _predict_from_image_handles_in_batches(
         model_: Model,
-        image_paths: Sequence[str],
+        image_handles: Union[Sequence[str], Sequence[Image.Image]],
         batch_size: int,
         preprocessing: Preprocessing
 ) -> NDFloat32Array:
@@ -68,11 +74,11 @@ def _predict_from_image_paths_in_batches(
     https://keras.io/api/models/model_training_apis/#predict-method
     """
     prediction_batches: List[Any] = []
-    for i in range(0, len(image_paths), batch_size):
-        path_batch = image_paths[i: i + batch_size]
+    for i in range(0, len(image_handles), batch_size):
+        handle_batch = image_handles[i: i + batch_size]
         image_batch = [
-            preprocess_image(Image.open(path), preprocessing)
-            for path in path_batch
+            preprocess_image(_load_pil_image(handle), preprocessing)
+            for handle in handle_batch
         ]
         prediction_batches.append(model_(np.array(image_batch)))
     predictions: NDFloat32Array = np.concatenate(prediction_batches, axis=0)
@@ -80,7 +86,7 @@ def _predict_from_image_paths_in_batches(
 
 
 def predict_images(
-        image_paths: Sequence[str],
+        image_handles: Union[Sequence[str], Sequence[Image.Image]],
         batch_size: int = 8,
         preprocessing: Preprocessing = Preprocessing.YAHOO,
         weights_path: Optional[str] = get_default_weights_path(),
@@ -88,12 +94,12 @@ def predict_images(
         alpha: float = 0.8
 ) -> List[float]:
     """
-    Pipeline from image paths to predicted NSFW probabilities.
+    Pipeline from image handles to predicted NSFW probabilities.
     Optionally generate and save the Grad-CAM plots.
     """
     _update_global_model_if_needed(weights_path)
-    predictions = _predict_from_image_paths_in_batches(
-        global_model, image_paths, batch_size, preprocessing
+    predictions = _predict_from_image_handles_in_batches(
+        global_model, image_handles, batch_size, preprocessing
     )
     nsfw_probabilities: List[float] = predictions[:, 1].tolist()
 
@@ -101,10 +107,11 @@ def predict_images(
         # TensorFlow will only be imported here.
         from ._inspection import make_and_save_nsfw_grad_cam
 
-        for image_path, grad_cam_path in zip(image_paths, grad_cam_paths):
+        for image_handle, grad_cam_path in zip(image_handles, grad_cam_paths):
+            assert isinstance(image_handle, (str, Image.Image))  # For mypy.
+            pil_image = _load_pil_image(image_handle)
             make_and_save_nsfw_grad_cam(
-                Image.open(image_path), preprocessing, global_model,
-                grad_cam_path, alpha
+                pil_image, preprocessing, global_model, grad_cam_path, alpha
             )
 
     return nsfw_probabilities
diff --git a/tests/_test.py b/tests/_test.py
index cc3b7fc..7909634 100644
--- a/tests/_test.py
+++ b/tests/_test.py
@@ -6,6 +6,7 @@
 from typing import Optional, Sequence
 
 from keras import backend as keras_backend
+from PIL import Image
 
 import opennsfw2 as n2
 
@@ -61,9 +62,14 @@ def test_predict_images_simple_preprocessing(self) -> None:
         self._assert(expected_probabilities, predicted_probabilities)
 
     def test_predict_image(self) -> None:
+        # From path.
         self.assertAlmostEqual(
             0.983, n2.predict_image(IMAGE_PATHS[1]), places=3
         )
+        # From PIL Image.
+        self.assertAlmostEqual(
+            0.983, n2.predict_image(Image.open(IMAGE_PATHS[1])), places=3
+        )
 
     def test_predict_video_frames(self) -> None:
         elapsed_seconds, nsfw_probabilities = n2.predict_video_frames(
diff --git a/tests/run_code_checks.sh b/tests/run_code_checks.sh
index 64c59c2..cd5328e 100755
--- a/tests/run_code_checks.sh
+++ b/tests/run_code_checks.sh
@@ -4,4 +4,4 @@ set -e
 find opennsfw2 -iname "*.py" | grep -v -e "__init__.py" | xargs -L 1 pylint --errors-only
 find opennsfw2 -iname "*.py" | grep -v -e "__init__.py" | xargs -L 1 pylint --exit-zero
 find opennsfw2 -iname "*.py" | grep -v -e "__init__.py" | xargs -L 1 mypy --strict --implicit-reexport --disable-error-code attr-defined --disable-error-code unused-ignore
-find tests -iname "*.py" | xargs -L 1 python3 -m unittest
\ No newline at end of file
+find tests -iname "*.py" | grep -v -e "__init__.py" | xargs -L 1 python3 -m unittest
\ No newline at end of file