TMPGEnc: Best setting for fast action movies

[Whoosh25]

From what I read on this forum, the older messages, this is a pretty good setting to use for fast-moving action videos, less jitter....right?...

TMPGENC Settings for NTSC VCD

Video Tab
Motion Search Precision: Highest Quality (very slow)

Advanced Tab
Sharpen Edge: 127 both horizontal & vertical. Keep field base UNCHECKED.

Quantize matrix Tab
Use Floating Point DCT.
UNCHECK No motion search for stll pictures part by half pixel
UNCHECK Soften block noise

Any other suggestions is appreciated...thanks.

[Whoosh25]

Anybody?!?

Don't tell me nobody tried this and everybody uses the default setting?

[MITSUI_1]

Well, with due respect, the answer really is that "It depends"

It depends on the quality of your source. If your source was "captured" footage with lots of dot crawl, then cranking up the "sharpen" is just going to accentuate that noise. MPEG1 has a tough time with sharp edges, so sharpening too much just causes the encoder to spend more bits getting the details right, with the result that macroblocks appear elsewhere.

Even if you decide you need to sharpen, whether or not to check on "field base" is determined not as a general preference, but depending on whether you're going from an interlaced source.

No one I've spoken with could honestly tell the difference between "high quality (slow)" and "highest quality (very slow)" motion search precision. Highest quality adds several hours to a 2 hour movie encoding session, however. So use "high" quality.

In terms of the other settings, your really just pecking around the real issue: If you've got a lot of motion, you've got to use a lot of bits. Period.

Make an xVCD at 1300 to 1500 bps video rate. If you're worried that your DVD player cannot handle it, make a short disk of test clips first. (find a high action trailer, and encode it at 1150bps, 1250bps, 1350bps, 1550bps, 1800bps, etc. See which one you like better).

Despite all you do, if there is still too much motion you will likely still see macroblocks. You can always buy the DVD at Best Buy!

[SatStorm]

When you use the sharp filter with progressive source/output, the results ain't that great, in my opinion....
Using the sharp filter with interlace source/de-interlace/interlace output, then for 352 X 288 (240) and bitrate about 1200, you get very good results with TMPGenc. New versions I believe are better with this. Of course, that means you use mpeg 2... Also, smaller gops helps, the more you use the sharp filter... I usually set it to 74, I'll try tonight to set it more and don't mark the field base
About the Highest and Higher modes, I have to say that on music videos (with much motion), I can see difference, on low bitrates. For bitrates beyond 1700, the quality looks the same....

Anyway, if you can use mpeg 2 and XSVCD, then give it a try.
Interlace output does the difference!

[Whoosh25]

Wow thanks guys...

My experiment with trying to match the quality of a purchased VCD movie. I ripped the exact movie from a DVD (lots of panning and movement) and is trying to match the same quality. I know the bought VCD is encoded with expensive equipment.

How about simply ...What would your personal TMPGENC settings be under those tabs for a Standard CBR NTSC VCD from a ripped DVD (Action-packed, Progressive)???

[Whoosh25]

I guess lots of people are wondering too!
I just checked that close to 400 individuals viewed this post...lol.

Come on VCD gods, share with us your supreme knowledge !

[vhelp]

Well, Whoosh25,...and others. . .

yes, i agree w/ mitsui_1, in regards to captured avi's.
If you up to much on the sharp, you WILL assentuate the
noise, hene more blocks. I experimented on this last
night. Many more blocks. Is it easy to find the best
"sharpness" setting. NO! But, what I do is encode a
short 5 to 10 second on a few scenes and try about 5
different settings. Then burn to CD. If using nero,
I set seconds to "0" and then burn, then, when I'm
ready to play on my dvd player, I set player to
"repeat-all" and watch for a couple of minutes (or more)
and then I make my final judgement. This is a laborous
process, but, once you figure out a good system for
this, its basically a piece of cake (sort of) and you
can do it quicker, 1,2,3...
I don't have my sharpenss at 127, but lower. Every dvd
player is different. So's every eye. But, if it's only
gonna be U watching it, then it doesn't matter, but if
others R, then you should try to come to a compromise.

I use tmpg for all my encodes. But, to be honest with CCE...
Over the weekend, I had a chance to play with CCE. It has
given me great trouble in the past, as well in the present.
But that's not what i'm after here. I just wanted to say, that in
the time I did get it to work (still trouble) I didn't
notice the sharpness being too bad. In fact, I felt it
didn't need to be tweaked, except for maybe a few. But, then
again, I only got to do just a handful of encode testings.
The thing kept crashing on my XP 1700+ system, which by the
way, isn't operating properly.

Another facter to consider when ajusting quality (or sharpness)
in your VCD's is the "field order". I've ben pretty using the
wrong field order in ALL my encodes, but I didn't really notice
them being off caues I used a de-interlace filter.
I think (IMO) that if you use the correct field order, you
don't have to de-interlace your video for VCD's. I've done
quite a few this way, and the quality has improved some.
In short, de-interlacing will add to some extent, distortion
because of the way it deals with each frame. Someone else here
more techy can explain this in detail (I hope) so you understand
why I say this.

In all my encodes, I've never used anything higher than NORMAL.
I will experiment on upping this today, and see how quality is
affected (on fast and slow scenes)

Currently, I test my VCDs with 1150 or 1250 bitrates, using
CQ_VBR and not CBR. So, basically, its an xVCD, though I call
it a VCD (generically speaking)

I've done some VCD clips last night from my Satalete. I've got
the issue settled with my sat. Anyways, I did a small cap of
"Star Trek E" and I liked what I saw, for a VCD. A little blurry,
but not as much as my prev. ones I did many, many months ago, which
by the way, turn me to my xSVCD's in stead. But, now that I have
a better grasp of VCD (thanks in part to SVCD) I am feeling more
curious about doing some VCDs. My only concirn is the quality.
How do I get the best quality IN a VCD from my sat. captures?
Only trial and error will tell. And, I'm getting closer. I'd
like to post a sample clip of what I did lastnight, just as a
teaser to my first attempt, for those curious what a sat capture
and encode will look like. Maybe. The sample was done with
min. settings. And, even that min. settings, I thought it was
pretty good for my first attempt.


I think as an experiment, to gaining better VCD quality, we ALL
should try an encoding test of a given DVD source, but use the
SAME source, then post results. Sort of like a contest, but there
are no winners. Just good endoing settings or methods to share.

I doubt that people will participate, but I thought that's a
good way to get the ball rolling on this VCD hunt!

-vhelp

[SatStorm]

OK, I finished this morning the test...

My source was an interlace DVB transmission. Resolution 544 X 576, mpeg 2, bitrate about 2500 - 5000 VBR. Movie lenght 111 min. It was that "6 days.." something movie with cloning Arnold, I don't remember the full title.

I used tmpgenc 2.51 plus and dvd2avi as always. I encoded to 352 X 288, mpeg 2, CQ_VBR 950 @50%, audio 128kb/s. I used the sharp filter about 127 both, unmark the field base and de-interlace the odd field.
The encoding took 3 hours exactly (!) for both audio/video with my 1700 Athlon XP. It use to take 3.30- 4.00 hours with 2.02 version for a same in lenght movie, but this depends from many parameters...
The result was a 838 mpeg-2 file. Now, I re-encode the movie with lower bitrate (870) to fit it on a 80min CD

Anyway, from what I saw the results are great! Very few blocks only on very extreme scenes!
Overall, the quality is much better VHS. It is worth for mainstream use!

On other tests with "standard" bitrates (about 1150), there were no blocks on ANY scene... Maybe my material ain't that difficult to encode (who knows!)
That unchecked "Keep field base" on the sharp fiter made a huge difference on the quality!
On PC monitor, it don't looked as well as files with marked the "keep field base". But my experience teach me to judge only on TV screens... So, I burn on a CD RW, the same clip, encode it twice with the same settings except the "keep folder field base". Amazing difference. No blocks with the unchecked version, only a few noticable blocks with the checked one....

I believe with 2.51 standard the results would be the same....

I don't know if those results would be the same with avis. I only test with DVB material.

So, the conclusion is: At last, xSVDs with low resolution and VCD bitrates, with tmpgenc 2.51 plus are equal (if not better ) VHS.
90-100 min per 80min CD on VHS quality ain't bad for mainstream use