On 8/24/07, Robert Tigelaar <[EMAIL PROTECTED]> wrote:
> First: How do I have to send mails to the mailing list so that they appear
> in the treeview instead of single entrys?

just reply to sender (using any decent [web] mail client that
preserves references)
should be enough :)

> Second: I like transcode a lot and I'll keep using it,

And this of course is good :)
Also, we want of course to improve transcode, sadly we suffer of lack
of manpower an
developer time :(

> movies I guess I'll need to use another tool like mencoder or ffmpeg, or
> would I just get the same desynced result with that?

It is possible, I'm not aware of features of other tools expliciting
addressing this
issue. On the other hand, their demuxing layer can be (and will likely
be) better
than ours.

> Oh well, learning is never a wrong thing right?

Never ever :)

[...]
> Fourth: What are the theories behind a/v desync? How and why does it happen?
> Maybe I can join the think thank, I'm not a coder but I can make up nice
> theories about how to solve stuff.

I've gathered some docs on the web and I can (hopefully) provide some pointers.
There isn't (AFAIK) a canon theory about this, it's a side effect of
how video encoding
works and how encoding it's done.

A quick example regarding the constant desync (shift IIRC the terminology).
Some video clips have the beginning with only video and no (= silence)
audio, maybe
during the opening credits/title. Of course the straightforward way to
obtain this is to insert
some silence audio frames (=there is audio data, but it
decodes/reproduces to plain silence).
A more clever (?) way is to cut audio silence frames -this also saves
some bits- and setup accordingly the timestamp of first valid (=not
silence) audio frame.

Maybe an example is useful

`dumb way', 2 frames of silence at the beginning of clip
timestamp (ms) | Video Data | Audio Data:
0                     |     yes         | yes, silence
40                   |     yes         | yes, silence
80                   |     yes         | yes
...

Physical layout: both audio and video tracks starts with LOGICAL frame
#0 and both have the same duration in frames. A/V shift is 0ms

`clever' way, 2 frames of silence at the beginning of clip:
timestamp (ms) | Video Data | Audio Data
0                     |     yes         | no!
40                   |     yes         | no!
80                   |     yes         | yes
...
Physical layout: audio track has N frames and starts with LOGICAL
frame #2, while
video track has N+2 frames and starts with LOGICAL frame #0. A/V shift is 80ms,
meaning that audio track MUST have a delay of 80ms respect to video track
to obtain proper sync.

Just a quick example, but I hope this can be useful to start
understanding those issues

Bests,

-- 
Francesco Romani // Ikitt

Reply via email to