You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Correct sequence should be 0xFC 0x94 0x2F 0xFF 0x80 and not 0xFC 0x94 0x2F 0x80
I've found this while investigating why FFMpeg couldn't extract the CC caption data of the samples provided by the test server on AWS (namely https://livesim.dashif.org/dash/vod/testpic_2s/cea608.mpd). This is not the sole reason for the lack of interoperability but at least seems to be the only issue out of the scope of FFmpeg.
It'd be great if this could be fixed and the public (affected) samples regenerated.
Thanks for the report and for investigating the issues. It seems indeed that I added the rbsp_trailing bits, but dropped the marker_bits at the same time. I agree that the SEI NAL units still come after the video NALUs. Will try to fix that.
This issue should now have been properly fixed and the assets updated.
The updated content has one line in italics, so it can be distinguished from the earlier version:
Dear all,
In Guidelines for Implementation: DASH-IF Interoperability Points v3.1 the following is stated regarding CEA-608/708 in SEI messages:
The description is shown below:
Unfortunately since 23cb15c#diff-3372455acaff98fed2743e2f609a073a02d36e8edc38bc43adbebc56a03dfef5R83 the cc_data is immediately followed by the
rbsp_trailing_bits
(the marker_bits (0xFF) might have been unintentionally removed).Considering the example of the last cc_data packet in one of the SEI elements:
Correct sequence should be 0xFC 0x94 0x2F 0xFF 0x80 and not 0xFC 0x94 0x2F 0x80
I've found this while investigating why FFMpeg couldn't extract the CC caption data of the samples provided by the test server on AWS (namely https://livesim.dashif.org/dash/vod/testpic_2s/cea608.mpd). This is not the sole reason for the lack of interoperability but at least seems to be the only issue out of the scope of FFmpeg.
It'd be great if this could be fixed and the public (affected) samples regenerated.
cc @tobbee
The text was updated successfully, but these errors were encountered: