Consider defining a Metadata OBU for private data

[cconcolato]

One could store SMPTE timecodes for example, or other metadata.

[sunghee-hwang]

Can we postpone this to v2?

[sunghee-hwang]

If we do have, then the below syntaxes of Timecode OBU payload would be good enough?

class timecode_obu() {
unsigned int (32) smpte_timecode_bcd;
}

'smpte_timecode_bcd' indicates HH:MM:SS:FF format and is represented using binary-coded decimal. Each variable represents in the range from 0-9 within the half-byte boundary. Where, H: hour, M: minute, S: second and F: frame.

For example, 15 hours 05 minutes 48 seconds 11th frame is indicated as follows:

• 1st byte (MSB): 0x15
• 2nd byte: 0x05
• 3rd byte: 0x48
• 4th byte (LSB): 0x10
  NOTE: It can represent up to 100 frames (0x00 ~ 0x99) per second.

[cconcolato]

There should be a way to carry generic metadata payloads, potentially defined elsewhere:

• TimeCodes as in AV1 https://aomedia.org/av1/specification/syntax/#metadata-timecode-syntax
• ITU-T T.35 messages (as in AV1, or other video codecs)

It could be simply reserving an OBU type, having an leb128 metadata type, as in AV1: https://aomedia.org/av1/specification/syntax/#general-metadata-obu-syntax

[sunghee-hwang]

Ok, I get it.
Then, I think that we have no action as long as there are enough number of reserved obu_types.
Am I right?

[cconcolato]

Let's discuss in meeting to see if we want that for v1 or v2.

[tdaede]

Let's leave this for v2. (Regardless if we reserve a OBU type for this or not, it does not change decoder behavior wrt skipping reserved OBUs).

[cconcolato]

For context, AOM is working on a potential new spec/registry of metadata, that could be useful to link with IAMF: https://aomediacodec.github.io/metadata-registry/