Store mp4 metadata at the beginning of the file when possible. (#15195)

comfy_api/latest/_input_impl/video_types.py

@@ -36,13 +36,15 @@ def get_open_write_kwargs(
     dest: str | io.BytesIO, container_format: str, to_format: str | None
 ) -> dict:
     """Get kwargs for writing a `VideoFromFile` to a file/stream with `av.open`"""
+    is_write_to_buffer = isinstance(dest, io.BytesIO)
+    is_mp4_file = not is_write_to_buffer and os.path.splitext(dest)[1].lower() == ".mp4"
+    movflags = "use_metadata_tags+faststart" if is_mp4_file else "use_metadata_tags"
     open_kwargs = {
         "mode": "w",
         # If isobmff, preserve custom metadata tags (workflow, prompt, extra_pnginfo)
-        "options": {"movflags": "use_metadata_tags"},
+        "options": {"movflags": movflags},
     }
 
-    is_write_to_buffer = isinstance(dest, io.BytesIO)
     if is_write_to_buffer:
         # Set output format explicitly, since it cannot be inferred from file extension
         if to_format == VideoContainer.AUTO:
@@ -103,7 +105,9 @@ def mp4_output_open_kwargs(path: str | io.BytesIO, format: VideoContainer, codec
         raise ValueError("Only MP4 format is supported for now")
     if codec != VideoCodec.AUTO and codec != VideoCodec.H264:
         raise ValueError("Only H264 codec is supported for now")
-    open_kwargs = {"mode": "w", "options": {"movflags": "use_metadata_tags"}}
+    # FFmpeg's faststart pass reopens the output by filename, so it cannot be used with file-like objects.
+    movflags = "use_metadata_tags+faststart" if isinstance(path, (str, os.PathLike)) else "use_metadata_tags"
+    open_kwargs = {"mode": "w", "options": {"movflags": movflags}}
     if isinstance(format, VideoContainer) and format != VideoContainer.AUTO:
         open_kwargs["format"] = format.value
     elif isinstance(path, io.BytesIO):

tests-unit/comfy_api_test/input_impl_test.py

@@ -36,6 +36,7 @@ def test_get_open_write_kwargs_filepath_no_format():
     kwargs_specific = get_open_write_kwargs("output.avi", "mp4", "avi")
     fail_msg = "Format should not be set for file paths (Specific)"
     assert "format" not in kwargs_specific, fail_msg
+    assert kwargs_specific["options"]["movflags"] == "use_metadata_tags"
 
 
 def test_get_open_write_kwargs_base_options_mode():
@@ -43,9 +44,9 @@ def test_get_open_write_kwargs_base_options_mode():
     kwargs = get_open_write_kwargs("output.mp4", "mp4", VideoContainer.AUTO)
     assert kwargs["mode"] == "w", "mode should be set to write"
 
-    fail_msg = "movflags should be set to preserve custom metadata tags"
+    fail_msg = "movflags should preserve custom metadata tags and enable faststart for MP4 files"
     assert "movflags" in kwargs["options"], fail_msg
-    assert kwargs["options"]["movflags"] == "use_metadata_tags", fail_msg
+    assert kwargs["options"]["movflags"] == "use_metadata_tags+faststart", fail_msg
 
 
 def test_get_open_write_kwargs_bytesio_auto_format():

tests-unit/comfy_api_test/video_types_test.py

@@ -258,6 +258,18 @@ def test_save_to_h264_crf_controls_quality(tmp_path):
     assert os.path.getsize(transcoded) < os.path.getsize(high_quality)
 
 
+def test_save_to_mp4_writes_metadata_before_media(video_components, tmp_path):
+    encoded = tmp_path / "encoded.mp4"
+    remuxed = tmp_path / "remuxed.mp4"
+
+    VideoFromComponents(video_components).save_to(str(encoded), metadata={"prompt": {"test": "value"}})
+    VideoFromFile(str(encoded)).save_to(str(remuxed), metadata={"prompt": {"test": "value"}})
+
+    for path in (encoded, remuxed):
+        data = path.read_bytes()
+        assert data.index(b"moov") < data.index(b"mdat")
+
+
 def create_transcode_source(
     width=64, height=64, frames=30, fps=30, audio_streams=1, undecodable_audio=0, rotation=False,
     container_format="mov", audio_codec="pcm_s16le",