[videophool]

The SEI data bytes are as follows:

0000000106001180009B4500000300000300009B450000030000403FFFFFFFC000309080080000030008800000000165

I have interpreted the bytes as follows:

0000000106
NAL Start code for SEI Message

00 11 80009B4500000300000300009B45000003000040
The first SEI message: Type = 00 (Buffering payload) size == 0x11 bytes (17 bytes).

3F FFFFFFC0 0030908008000003000880
This is presumably the second SEI message. The type is 3F , which is not a defined SEI message. From the spec, I understand that this should be parsed as a ‘reserved_sei_message’. However, note the size bytes of the SEI FFFFFFC0. This means that the SEI payload size is 3x255 + C0, or 957 bytes (0x3BD). However, you can see that there are only 11 bytes, followed by the Slice NAL start code (00 00 00 01 65).

My first thought was that this was a stream corruption (a 3rd party analyzer is also complaining about the SEI). However, I have two different streams (out of tens of thousands of encodes) that have the exact same data bytes following the buffering period SEI. The problem is not random corruption. Seems to be a problem with the encoder.

Any thoughts or help would be greatly appreciated.