diff --git a/CHANGELOG.rst b/CHANGELOG.rst
index ce37ff4f9..2a8d815b7 100644
--- a/CHANGELOG.rst
+++ b/CHANGELOG.rst
@@ -39,9 +39,12 @@ Features:
 - Support reusing the thread's current CUDA context via a ``current_ctx`` flag on ``CudaContext`` and ``VideoFrame.from_dlpack``, for interop with libraries like PyTorch that initialize CUDA first by :gh-user:`Yozer` (:pr:`2339`).
 - ``VideoFrame.from_dlpack`` no longer requires restating ``primary_ctx``/``current_ctx`` when passing an explicit ``cuda_context``; the flags are only validated when explicitly given by :gh-user:`WyattBlue`.
 - Support passing an explicit CUDA stream to FFmpeg CUDA operations, including NVENC input and output, via a ``cuda_stream`` parameter on ``CudaContext``; currently limited to logical CUDA device 0 by :gh-user:`Yozer` (:pr:`2360`).
+- ``VideoFrame.save`` now forwards keyword arguments to the encoder, letting callers trade file size for speed (e.g. ``pred="none"`` or ``compression_level=1`` for PNG, ``qscale=2`` for JPG) by :gh-user:`WyattBlue`.
 
 Fixes:
 
+- Fix ``VideoFrame.save`` raising ``AttributeError`` when given a ``Path`` rather than a ``str`` by :gh-user:`WyattBlue`.
+
 - Prevent crashes and corrupted output when structural codec properties are changed after an output stream has been opened by :gh-user:`WyattBlue`, reported by :gh-user:`oakaigh` (:issue:`2232`).
 - Fix a crash when using a stream that has no ``CodecContext`` (a demuxed stream with no available decoder, such as one from a truncated file, or a stream created by ``add_mux_stream``); decoding now raises ``DecoderNotFoundError``, encoding now raises ``EncoderNotFoundError``, and ``BitStreamFilterContext`` accepts such a stream as ``out_stream`` by :gh-user:`WyattBlue`, reported by :gh-user:`justinrmiller` (:issue:`2344`).
 
diff --git a/av/video/frame.pxd b/av/video/frame.pxd
index 66926058d..7f93c99ec 100644
--- a/av/video/frame.pxd
+++ b/av/video/frame.pxd
@@ -25,6 +25,5 @@ cdef class VideoFrame(Frame):
     cdef readonly int _device_id
     cdef _init(self, lib.AVPixelFormat format, unsigned int width, unsigned int height)
     cdef _init_user_attributes(self)
-    cpdef save(self, object filepath)
 
 cdef VideoFrame alloc_video_frame()
diff --git a/av/video/frame.py b/av/video/frame.py
index 74964107b..0d22c481c 100644
--- a/av/video/frame.py
+++ b/av/video/frame.py
@@ -773,17 +773,21 @@ def to_rgb(self, **kwargs):
         """
         return self.reformat(format="rgb24", **kwargs)
 
-    @cython.ccall
-    def save(self, filepath: object):
+    def save(self, filepath: object, **options):
         """Save a VideoFrame as a JPG or PNG.
 
         :param filepath: str | Path
+        :param \\**options: Encoder options, e.g. ``pred="none"`` or
+            ``compression_level=1`` for PNG, ``qscale=2`` for JPG. Values are
+            coerced to ``str``. The PNG defaults favor file size over speed;
+            ``pred="none"`` is roughly 3x faster and 2x larger.
         """
         is_jpg: cython.bint
+        name: str = str(filepath)
 
-        if filepath.endswith(".png"):
+        if name.endswith(".png"):
             is_jpg = False
-        elif filepath.endswith(".jpg") or filepath.endswith(".jpeg"):
+        elif name.endswith(".jpg") or name.endswith(".jpeg"):
             is_jpg = True
         else:
             raise ValueError("filepath must end with png or jpg.")
@@ -794,7 +798,11 @@ def save(self, filepath: object):
         from av.container.core import open
 
         with open(filepath, "w", options={"update": "1"}) as output:
-            output_stream = output.add_stream(encoder, pix_fmt=pix_fmt)
+            output_stream = output.add_stream(
+                encoder,
+                pix_fmt=pix_fmt,
+                options={k: str(v) for k, v in options.items()},
+            )
             output_stream.width = self.width
             output_stream.height = self.height
 
diff --git a/av/video/frame.pyi b/av/video/frame.pyi
index f07a8079f..b2240ab9c 100644
--- a/av/video/frame.pyi
+++ b/av/video/frame.pyi
@@ -85,7 +85,7 @@ class VideoFrame(Frame):
         threads: int | None = None,
     ) -> VideoFrame: ...
     def to_rgb(self, **kwargs: Any) -> VideoFrame: ...
-    def save(self, filepath: str | Path) -> None: ...
+    def save(self, filepath: str | Path, **options: Any) -> None: ...
     def to_image(self, **kwargs): ...
     def to_ndarray(
         self, channel_last: bool = False, **kwargs: Any
diff --git a/tests/test_videoframe.py b/tests/test_videoframe.py
index acd3f53ab..36a1d5386 100644
--- a/tests/test_videoframe.py
+++ b/tests/test_videoframe.py
@@ -1379,3 +1379,21 @@ def test_reformat_pixel_format_align() -> None:
         result = frame_rgb.to_ndarray()
         assert result.shape == expected_rgb.shape
         assert numpy.abs(result.astype(int) - expected_rgb.astype(int)).max() <= 1
+
+
+def test_save_options(tmp_path) -> None:
+    y, x = numpy.mgrid[0:240, 0:320]
+    array = numpy.dstack([x % 256, y % 256, (x + y) % 256]).astype(numpy.uint8)
+    frame = VideoFrame.from_ndarray(array, format="rgb24")
+
+    default = tmp_path / "default.png"
+    unfiltered = tmp_path / "unfiltered.png"
+    frame.save(default)
+    frame.save(unfiltered, pred="none")
+
+    # pred="none" skips the row filter: much faster, much bigger.
+    assert unfiltered.stat().st_size > 4 * default.stat().st_size
+
+    # Non-str values are coerced.
+    frame.save(tmp_path / "q.jpg", qscale=2)
+    assert (tmp_path / "q.jpg").stat().st_size > 0