see w3c/dapt#18

We're seeing increasing use of Stellar for multi-talent voice over with mixing with original voice audio, requiring fade to be set per event.

The requirements should be adjusted to account for Process Step 4 being optional for a 'dubbing' workflow where original M&E is not available, or where it is desirable to mix over the original voice.

Given that both PS4 and PS5 now apply to Dubbing and AD, it would be good to make them sequential. However, considering that adjusting words is more natural right after recording and that adding mixing instruction can be seen as the final step, I would do PS3->PS5->PS4 (and renumber them).

In §2.7 the second bullet seems to combine two things, an amount and a rate, so appears confusing. Separate out the requirements of amount, and rate, of fade.

Received as direct email feedback to me from Matt Simpson of ITV.

From @nigelmegitt :

In ADPT, the requirement for intervals was distinct from the requirements their contents - any reason not to do the same here?

@nigelmegitt said:

Is "event" the best name? Do nested events need a different term?

It would be good if the diagram and the tables are accessible to screen readers. From my first checks, there seem to be some difficulties.

from transition request, it was pointed that there is no text as Dubbing and Audio description Profiles of TTML2 within this reqs document. it's worth to have some line in abstract or overview.

@nigelmegitt @gkatsev @cconcolato how do you think?

Pan:
Although in all my experience of AD over the years I have never seen PAN used in the wild, the PAN value should be presented in the same way as fade - a set of desired PAN values at time offsets from the incue, so that the description can move. (i.e. the descriptive construct can be the same data type as Fade.)

5.1 (or object based audio):
Indeed, I would love a definition of 3d placement an interpolation between points, and a definition of how to fade the original track(s). Maybe there is something in 'Object based audio' we can draw on?

(note that 'stereo mixing' and '5.1 mixing' data may need to be separately stated, although there could be a default way of deriving one from the other?).

Gain:
In Stellar, all audio 'editing' is non-destructive. So we allow for a single gain for each recording. We also have a mute flag.
Multiple audios may be associated with a time period, and they may or may not be muted.

It is quite important to understand that:
Fade apples to the original audio, and so there is one per time period/script box.
Gain/Placement/Mute applies to an audio recording, of which there may be multiple per time period/script box, all with different offsets and trims. (e.g. a description may be made up of the first half of one take, and the second half of another take, and there may be three other muted takes present).

It is complex to implement a fade per audio where audios overlap/are close, hence the split of logical location of the data to ensure only one fade per 'script box'/time period.

Note that I would not be averse to inclusion of 'audio processing instructions' - e.g. in Stellar mixing, we support complex gain profiles for audio recordings, and complex compression/expansion/noise gate, etc. which we will have as a full implementation in support of de-essing, loudness control and noise reduction within a few months. The file format must be able to support these even if as private metadata against an audio if it is to be of use to us as a preparation format.

ADPT had [media-accessibility-reqs] info in the table. It should be added.