Best codec

[Blinkr]

I know this is probably a "it depends" type of question, but I thought I would try it anyway.

What is the best codec to use when capturing to avi's or mpeg2's??? I captured some video to an avi with no compression (using Stoik) & my Pinnacle Studio 8 wouldn't open it. I then captured it using one of the listed codec's in it's capture setup & it would then recognize it. It lists about 5 or 6 codecs while my capture app (Stoik) list alot more.

I would like to get some suggestions to make sure I'm getting the best quality capture.

Thanks!!!

[hrlslcbr]

If you want 'the best possible quality', don't use any codecs, or use a lossless codec (huffyuv, MSU, x264, etc).

[BJ_M]

x264 isnt lossless per say, normally ....

msu is slow encoding but very good quality ,,

huffyuv and Lagarith and mjpeg or Alparysoft are all possible contenders

[hrlslcbr]

Yes. x264 CAN be lossless (and very good too).

[daamon]

I use Adobe Premiere and it's geared towards DV AVI - maybe Pinnacle is the same?

There are a number of codecs for DV AVI - Do search in the "Tools" for "DV codec" to get the ones listed on this site. There is also a Microsoft DV codec, but some say it's not so good...

DV AVI is compressed, but at 25Mbps - 3x that of DVD so still plenty of quality compared to DVD.

As for MPEG2 codecs for encoding - I'm not sure on that. I think that, when you buy your encoder you pay for the license for a good codec to come with it. Like I say, I'm not too sure...

[hrlslcbr]

The only problem with DV is the 4:1:1 sampling if you capture NTSC videos. If you capture PAL, it's a great format.
But as BJ_M said, try using one of the codecs he suggested and you'll get the best possible quality.

[BJ_M]

Yes. x264 CAN be lossless (and very good too).

it is near lossless - not true lossless , and i said normally anyhow ...
and since it requires more cpu power to compress on the fly than other choices - not the best choice for capturing ....

if you do a A minus B compare of x264 vs. 4:4:4 original , you will quickly see it is not lossless

[hrlslcbr]

Yes. x264 CAN be lossless (and very good too).

it is near lossless - not true lossless , and i said normally anyhow ...
and since it requires more cpu power to compress on the fly than other choices - not the best choice for capturing ....

if you do a A minus B compare of x264 vs. 4:4:4 original , you will quickly see it is not lossless

Yes, of course. With the current version of x264 you can't compress 4:4:4, but it is lossless just like huffyuv is, for 'normal' sources (those are the type of sources we mortals can get... ). But I doubt that his captures will be more than 4:2:2 anyway.
Back to the topic. If you really want good quality captures and FAST (<--noticed that?) encoding, use Huffyuv or Alparysoft, which are free.

[vhelp]

My take is like this ...

If you want (as much (or closely) as possible) lossless video,
(assuming captures) then you want to *not* incorporate any
filters in your editing -to- final MPEG encoding process.

Ok. What I mean here ...

Say you capture with HUFFY, (a near lossless codec) then, when
you bring it in to edit it (say in vdub) to say, cut out some
commericals and that's it.., then you next step would be to
import the final edited source file into your MPEG encoder.
.
Inside your MPEG encoder, do not activate any type of color
filtering. Just encode the source.

This (the above) will give you lossless finalized MPEG's.
I say "lossless" in the sense, that after the MPEG compression,
it is lossless. If your bitrate is low, as in most VBR setups,
this is out the window, because now, we're on another path in
the form of macroblocks (pixelation) on account of the low bitrate.
That is my opinion.
.
Anyways..
.
The reason I say it is lossless (after MPEG compression) is
because that captured source (AVI) is in YUV color space, which
is what MPEG is in also. And, depending on the MPEG encoder used,
if it's color space conversion method is properly tuned (<--IMO)
then, as long as you *DO NOT* alter color during the editing and
MPEG encoding stage, all the MPEG detail will still be there,
hence the quality, hence the lossless.. or lossless feel

I've said this before, and I'll say it again.. when you capture a
given source with any codec, the formula should always be OUTPUT=INPUT.
That is, once the source is processed to MPEG, this OUTPUT should be
equal to the capture INPUT (aka, AVI source file) is all. In other
words, the color space of the OUTPUT should look identical to the INPUT
source.

To demonstrate what I'm on about here, lets use my favorate Holy Grail,
the ADVC-100 device as an example below.
.
The image below is a split view (two different images cut in half and then
stitched together to look like one complete image)

Step 1
Capture antenna signal source tv program (What I like about you) using
the ADVC-100 DV capture. It captures in YUV411 color space.

Step 2
Using vdubMOD, import the captured DV source file, (add a little VHELP
logo signiture to the source) and frameserve it into TMPGenc

Step 3
Using TMPGenc MPEG encoder, and encode This DV avi source file using the
following params of: CBR 9000 Bitrate.

Step 4
Still keeping the first vdubMOD open w/ the original DV source file, I
open another vdubMOD and import the encoded MPEG-2 file so that both
source and MPEG-2 scene's frame position can be saved as a BMP file.

Step 5
Next, using a stitching app I developed (for stitching two images together
for this purpose) I open both images, orignal and MPEG-2 BMP's and cut
out the original's rightside of the frame, and then cut out the MPEG-2 's
leftside of the frame. Then, I "stitch" the two 1/2 frames together to form
one whole frame. The purpose/advantage of this tool, is to demonstrate
any quality difference. Color space is one aspect to measure, in this
case scenario, hence the demonstration.

Now, viewing the image below, one can see with the naked eye that there
is *no* difference in quality, colorspace 'wise. About the only lossless
one might notice, is the macroblock (pixelation) in the scene, on account
of the source being too noisy (my antenna) and even CBR 9000 bitrate in
some cases, might not be enough. At quick glance, I did not notice any
pixelation. So, in this test case scenario, I would say that this test
proved lossless

Do you all agree ??



-vhelp 3430

[hrlslcbr]

as long as you *DO NOT* alter color during the editing and
MPEG encoding stage, all the MPEG detail will still be there,
hence the quality, hence the lossless.. or lossless feel
-vhelp 3430

I agree that any colour space conversions should be avoided to preserve the original quality, but TMPGEnc always converts to RGB to input video and then back to YUV (MPEG), so it isn't a lossless conversion.

So, Blinkr; the answer is...it depends... . If you want to preserve quality when capturing videos, use a lossless codec, and don't convert between colour spaces when not needed in your editing process.
Do we all agree?