fix: derive $VIDEO_NAME from the path of URL inputs

scenedetect/_cli/context.py

@@ -537,6 +537,11 @@ def _open_video_stream(
                     frame_rate=frame_rate,
                     backend=backend,
                 )
+            if not self.video_stream.name:
+                logger.warning(
+                    "Could not determine a video name from the input path, so $VIDEO_NAME will be"
+                    " empty in output filenames."
+                )
             duration = self.video_stream.duration
             duration_str = f"{duration} ({duration.frame_num} frames)" if duration else "unknown"
             rate = self.video_stream.frame_rate

scenedetect/platform.py

@@ -24,6 +24,7 @@
 import string
 import subprocess
 import sys
+import urllib.parse
 
 import cv2
 
@@ -129,11 +130,50 @@ def _get_cv2_param(param_name: str) -> int | None:
 ##
 
 
+def is_url(path: StrPath) -> bool:
+    """True if `path` looks like a URL or network stream (e.g. https://..., rtsp://...)."""
+    return "://" in os.fspath(path)
+
+
 def get_file_name(file_path: StrPath, include_extension: bool = True) -> str:
     """Return the file name that `file_path` refers to, optionally removing the extension.
 
-    E.g. /tmp/foo.bar -> foo"""
-    file_name = os.path.basename(os.fspath(file_path))
+    E.g. /tmp/foo.bar -> foo
+
+    URL and URI-like inputs:
+        Inputs containing "://" are treated as URLs (see :func:`is_url`) and handled by their
+        URL components rather than as paths, since a path separator inside a query string or
+        fragment is not part of the name. The name is the last non-empty segment of the path
+        component, percent-decoded, e.g. https://example.com/videos/clip.mp4?token=1.2 -> clip.
+
+        When the path component is empty there is no segment to use, so the host becomes the
+        name, e.g. rtsp://192.168.0.1:554 -> 192.168.0.1_554. Userinfo (credentials) is never
+        part of the result, and the port is joined with "_" because ":" is not valid in a file
+        name on Windows. Hosts are returned in full even when `include_extension` is False, as
+        the last "." in a host is a domain separator and not an extension.
+
+    Arguments:
+        file_path: Path, or URL, to get the name of.
+        include_extension: If False, remove the extension (the last "." and everything after
+            it) from the name. Has no effect on names derived from a URL host.
+
+    Returns:
+        The name, or an empty string if none can be derived (e.g. a path that is only
+        separators, or a URL with neither a path nor a host).
+    """
+    file_path = os.fspath(file_path)
+    if is_url(file_path):
+        parsed = urllib.parse.urlsplit(file_path)
+        segments = [urllib.parse.unquote(seg) for seg in parsed.path.split("/") if seg]
+        if not segments:
+            # No path to take a name from, so fall back to the host (never the credentials). A
+            # host has no extension to remove: stripping after the last "." would turn
+            # example.com into example.
+            host = parsed.hostname or ""
+            return f"{host}_{parsed.port}" if host and parsed.port else host
+        file_name = segments[-1]
+    else:
+        file_name = os.path.basename(file_path)
     if not include_extension:
         last_dot_pos = file_name.rfind(".")
         if last_dot_pos >= 0:

tests/test_platform.py

@@ -19,7 +19,7 @@
 
 import pytest
 
-from scenedetect.platform import CommandTooLong, invoke_command
+from scenedetect.platform import CommandTooLong, get_file_name, invoke_command, is_url
 
 
 def test_invoke_command():
@@ -38,3 +38,50 @@ def test_long_command():
     if platform.system() == "Windows":
         with pytest.raises(CommandTooLong):
             invoke_command(["x" * 2**15])
+
+
+@pytest.mark.parametrize(
+    ("path", "expected"),
+    [
+        ("/tmp/foo.bar", "foo"),
+        ("foo.bar", "foo"),
+        ("foo", "foo"),
+        ("https://example.com/videos/clip.mp4", "clip"),
+        # Query strings and fragments are not part of the name, even if they contain a "." or "/".
+        ("https://example.com/videos/clip.mp4?token=1.2&dir=a/b", "clip"),
+        ("https://example.com/videos/clip.mp4#t=10", "clip"),
+        ("https://example.com/videos/clip.mp4/", "clip"),
+        ("https://example.com/videos/my%20clip.mp4", "my clip"),
+        ("rtsp://user:pass@192.168.0.1:554/live/stream1", "stream1"),
+        # No path component: fall back to the host, which must stay a valid file name.
+        ("https://example.com", "example.com"),
+        ("https://example.com/", "example.com"),
+        ("rtsp://192.168.0.1:554", "192.168.0.1_554"),
+        ("rtsp://user:pass@192.168.0.1:554", "192.168.0.1_554"),
+        # Neither a path nor a host to derive a name from.
+        ("rtsp://", ""),
+        ("rtsp://:554", ""),
+    ],
+)
+def test_get_file_name_without_extension(path: str, expected: str):
+    assert get_file_name(path, include_extension=False) == expected
+
+
+@pytest.mark.parametrize(
+    ("path", "expected"),
+    [
+        ("/tmp/foo.bar", "foo.bar"),
+        ("https://example.com/videos/clip.mp4?token=1", "clip.mp4"),
+        ("https://example.com/", "example.com"),
+    ],
+)
+def test_get_file_name_with_extension(path: str, expected: str):
+    assert get_file_name(path, include_extension=True) == expected
+
+
+def test_is_url():
+    assert is_url("https://example.com/clip.mp4")
+    assert is_url("rtsp://192.168.0.1:554/live")
+    assert not is_url("/tmp/clip.mp4")
+    assert not is_url(r"C:\videos\clip.mp4")
+    assert not is_url("img_%04d.jpg")

website/pages/changelog.md

@@ -5,6 +5,7 @@
 
 ### Unreleased
 
+ - [bugfix] `$VIDEO_NAME` is now derived from the path component of URL inputs, ignoring the query string and fragment (e.g. `https://host/clip.mp4?token=1.2` -> `clip`). URLs with no path use the host. Previously the query string leaked into output filenames, which fails on Windows. [#197](https://github.com/Breakthrough/PySceneDetect/issues/197)
  - [bugfix] OTIO exports use the full source duration when provided, including from the CLI, instead of declaring a fixed 1,980-frame duration. The available range remains unspecified when the duration is unknown. Scene source ranges are unchanged. [#580](https://github.com/Breakthrough/PySceneDetect/pull/580)
 
 ### PySceneDetect 0.7.2 (October 11, 2026)