[MPlayer-users] RFC: docs update for "how to create a high quality DVD rip"

[D Richard Felker III]

On Sun, Jun 06, 2004 at 07:59:50PM -0400, Jason Tackaberry wrote:
> On Sun, 2004-06-06 at 19:46 -0400, D Richard Felker III wrote:
> > It doesn't work with inverse telecine or other filters that drop
> > frames (or add them), it doesn't work with field-coded DVDs (and it's
> > difficult to tell in advance which ones are field-coded), and it
> > causes A/V desync. Also it has no useful purpose.
> 
> I wonder if maybe I misunderstand how 3-pass works.  I figured the first
> pass was to compress the audio with VBR.  The next two passes work just
> like regular 2-pass, because now the (average) bitrate of the audio is
> known.  Given that, I don't see why filters that add/drop frames don't
> work with 3-pass.
> 
> Require enlightenment ...

Nope, the first (frameno) pass does the a/v sync (choosing which
frames to drop or duplicate) which can't be done correctly without
decoding and filtering them. And it doesn't decode them. So it's
broken.

If you want to do the same sort of thing correctly, you should encode
the audio separately with lame, then mux it with -oac copy and
-audiofile during the normal 2 passes. This is actually the best way
since bugs in mencoder cause slight a/v desync with modern versions of
lame due to lack of buffer measurement.

> > Um, let's see. 900/96 video/audio bitrate, or 804/192? I'll take the
> > former any day! It's an absolutely HUGE difference! For me. I
> 
> A good part of the movie experience for me is the sound.  I put a lot of
> money into my home theatre setup, and when I hear crap in my rear
> speakers because my receiver is trying to find a channel in the high
> frequencies that has been compressed to hell, I go batty.  It's really
> annoying. :)

Then disable that "feature"...

> > Otherwise (whenever you have a size constraint, no matter how big)
> > most of your recommendations are the exact opposite of what helps, and
> > stupid users will try them anyway...
> 
> I wonder if "most" is true.  Out of all the points in that text, I think
> the only ones that will hurt a size constraint is "don't scale" and
> "don't transcode audio."

Also: don't denoise.

> Speaking of "don't scale," what's your opinion on cropping to multiples
> of 16, or cropping to the exact rectangle and then scaling up to a
> multiple of 16?

Never scale up. Always scale down or not at all. Personally my pick
depends on whether the original was a movie or made-for-tv. I always
try to avoid vertical scaling when the original was made for tv (since
the line sampling corresponds to the way the content was originally
recorded) but that's pretty irrelevant when the original is film.

> > And several hundred more gigs for backups? :) Storing your only copies
> > on hd is not very smart...
> 
> They're not my only copies.  I have perfectly good backups sitting on a
> shelf in the other room: the actual DVDs. :)

And it takes months to restore them (reencode).

> > I _know_ you'll see them if you leave black borders, but insane
> > bitrate might mostly compensate, as least with qmin=1. With unaligned
> > dimensions I haven't tested, but in theory it should have a similar
> > effect.
> 
> I guess that makes sense.  Lower bitrates means you have a lot less room
> for mistakes in cropping, scaling, etc.

Yes, or conversely using insane bitrates lets you get by with being
sloppy. And I don't like sloppiness.

Rich