Open
Description
Description
Currently, the init_video_stream method in the PyAV plugin converts float fps to a Fraction. However, for some common fps numbers, this results in an OverflowError: value too large to convert to int.
Steps to Reproduce
The full code to reproduce the issue is as follows:
writer = iio.imopen("<bytes>", 'w', extension='.mp4', plugin='pyav')
writer.init_video_stream('h264', fps=29.97)---------------------------------------------------------------------------
OverflowError Traceback (most recent call last)
Cell In[36], line 1
----> 1 writer.init_video_stream('h264', fps=29.97)
File ~/miniconda3/envs/test/lib/python3.9/site-packages/imageio/plugins/pyav.py:886, in PyAVPlugin.init_video_stream(self, codec, fps, pixel_format, max_keyframe_interval, force_keyframes)
850 """Initialize a new video stream.
851
852 This function adds a new video stream to the ImageResource using the
(...)
882
883 """
885 fps = Fraction.from_float(fps)
--> 886 stream = self._container.add_stream(codec, fps)
887 stream.time_base = Fraction(1 / fps).limit_denominator(int(2**16 - 1))
888 if pixel_format is not None:
File av/container/output.pyx:93, in av.container.output.OutputContainer.add_stream()
File av/utils.pyx:51, in av.utils.to_avrational()
OverflowError: value too large to convert to int
Cause
This issue arises because float number is converted to a Fraction with very large numerator and denominator. For example, Fraction.from_float(29.97) returns Fraction(1054475631502295, 35184372088832).
Impact
This is difficult for me to work around since I am trying to maintain the original fps of the input file, as in
fps = reader.metadata()["fps"]Metadata
Metadata
Assignees
Labels
No labels