PDA

View Full Version : After some 60 movies made, I am asking: do we need multi-pass???


jmartina
24th February 2003, 23:53
First of all, I am talkin about DVD CCE enconding.

After some 60 backup copies of my movies made, I wonder
is it realy necessary to have multi-pass encoding.

I know that in theory that picture should be better as more passes
there are, but????

I have made some very big movies using one pass VBR, and
multi-pass VBR, movies ligr LOTR, Star Wars EP1, Star Wars EP2 etc.

To be honest on my 72cm Sony WEGA I can't spot the difference.
Picture is practicly identical to original, and if there are some
problematic scenes (darker mostly), they are problematic even after 5 passes.

Sometimes, I spot scene that is not so perfect, and I said A-HA there it is, but
then I put original DVD, go to same scene and I see it is same on the original.

I have also made some tests with DV. I have recorded some dark scenes using
my camera, some fast like car passing by quickly, and after encoding the result
is same. I am wasting my and my computers time using multi-pass.

So is there anyone that can say that there are some difference??

atreides93
25th February 2003, 08:05
I haven't done a comparison so I can't say that's a reason I do multipass, so the only real reason I use multipass is so I know exactly what the output will be regardless of what I'm feeding it. I know with constant Q VBR I can approximate the final output but I prefer using up every last byte I can fit on the DVD-R

MickeyNBK
25th February 2003, 11:59
Beauty is in the eye of the beholder. If one pass looks good enough for you, then yes you are wasting your time doing a multipass. Just made them to suit yourself.

fasttimes
26th February 2003, 03:20
:( Multi-pass is not mainly about quality. It's more about predictable file size. That is what 1-pass VPR cannot offer. Understand?

Ntsita
26th February 2003, 08:14
Originally posted by fasttimes
... That is what 1-pass VPR cannot offer...
With 'Bach One Pass VBR' tool is possible to predict the file size in this modality.;)

jmartina
27th February 2003, 00:04
fasttimes, I have always managed to produce .m2v file that fits on DVD
with 50-60MB left free using any bitrate calculator, and using only
one pass VBR.

I am not newbie here, and after many movies that I have made using BOTH
one-pass VBR and multi-pass VBR, I have noticed that there are no difference
in resultant movie even after nine-pass VBR (of course I have tried 9-pass
only with 5 minutes files, but complicated scenes).

So what I am asking here is not why multi-pass as a begginer's question,
but because I have spoted no difference in quality, what are YOURS expiriences.

MickeyNBK says it's in the eye of the beholder. Ok, I asked couple of my friends
to tell are they looking original movie, or re-encoded. They can't say.
And those movies were lareg ones - LOTR, StarWars Episode 1 & 2, Gladiator...

What can you see MickeyNBK???


Have you ever tried to use just one pass (say on DVD-RW), and check against
multi-pass on regular TV?

Why wait for 4 or six hours, when it can be done in 2 or so?

Again this is NOT newbie question.
This am asking as someone who have expirience, and who wasted days and days
using different encoders, and options within them.

So this question is for my brothers in arms.

atreides93
27th February 2003, 00:59
Ok the reason you don't see a difference is you have sufficient space and therefore a high enough bitrate, that there is no real difference between multiple passes and 1 pass VBR.

But, if you lowered the bitrate a lot to try to squeeze more video in, then you would start to notice differences because the multiple passes method would be capable of using those fewer bits to achieve more.

So you see...if your bitrate is high enough you might as well just do 1 pass VBR.

I read an article where a guy did a comparison like you did, he tried it at 3 megabits/s and noticed no difference between multipass and single pass...then he lowered it to 2 megabits/second and noticed significant differences in quality.

so there you go. enjoy

Prosper
27th February 2003, 22:40
I usually use multi pass, but set the # of passes to 1 (one for the vaf, and 1 for the m2v, so 2 passes altogether) - looks the same to my eye as a 4 pass, and file size accuracy is spot on.

jmartina
28th February 2003, 00:12
Atreides93, since 2 Mb/s is aprox 4 hours movie on DVD, and there is just a few
such movies, is it right conclusion that single-pass VBR is enough???

Of course there are lot's of people here that using CCE for VCD, but I am refering
to DVD only.

symonjfox
9th July 2003, 20:16
I used CCE multipass for SVCD purpose for years. I really think it helps on low bitrate conditions.

Keep in mind that SVCD has lot of limits and so it's not easy to allocate the bitrate as best as you can in 1 pass.
For example an average bitrate of 1570 kbs with min 300 and max 2500.
If you try using 1 pass VBR, the result is not so good. If you perform AT LEAST 2 passes, the overall quality is better (well, hi complexity scenes are limited to 2500 kbs so they won't look so good).

For DVD purpose there aren't those limits, so you can perform a min 300, avg 2000, max 9000 encode in 2 passes and keeping very very good results. There's no need to perform 9 passes (everybody will tell you that after the 4th pass there are very very few changes in the bitrate allocation).

screw
10th July 2003, 12:13
I am using the same method Prosper mentioned (AFAIK, it comes from Robshot guide) - first using one pass VBR create VAF, and than switch to multipass VBR, set passes to 1 and in "Advanced settings", or "Bit allocation" (in newer CCE versions) modes set your bitrates to get exact file size. So this requires just two passes.

Since I am doing DVD-s only, cannot comment about SVCD-s.

Kilyan
19th July 2003, 00:17
Hi!

I think using multipass, helps the coder to place only the necessary number of bits to a block, and use the other for more complicated ones, which will have a bigger matrice.
For seeing the difference you should try with the quantization scaling 1-100, and changing the inter and intra matrices, BTW why do CCE use an inter matrice which has different values from 16? The inter matrice is used for difference error coding, nothing to do with horizontal and vertical frequencies.

malboroman
24th July 2003, 23:37
Hi y'all,

I think I might be able to make a comment on this soon... I am currently re-encoding "Once Upon A Time In America" which is 4 hours and 47 min's long, so I have to use an incredibly low bitrate to fit this on a DVD-R. I will post my findings as soon as I'm done with some testing. I'm trying the RoBa method and 4-pass VBR.

My goal is to reauthor this 'resized' .m2v back into the movie and burn on DVD-R, and in the process add my own subs.(I have the R1 version and I want to add Dutch subs for... actually I have no reasons, but I just wanna do it to see if I can do it:))

Malboroman

symonjfox
24th July 2003, 23:48
1 tip I can give you is to RESIZE your video to 704*576 (cropped D1, 100% DVD compilant, allows 4:3 and 16:9 encoding), and also LETTERBOX 8 pixel from each border (that will never be visible on a TV screen).
Also a CPU=6 in MPEG2DEC3 1.08 will give you better compressibility (it will soft a little bit the image, but IMHO quality is still great) and for 4 hrs of video in 1 DVD this will be a good compromise.

Off course doing 4 passes in a such long file will help to redistribuite the bitrate to keeping costant the quality.

Another thing: CUT or encode @ low bitrate the CREDITS.

For example:LoadPlugin("C:\Programmi\avisynth 2.5\plugins\mpeg2dec3.dll")
Mpeg2Source("VTS__09_P01A1.16~9_1.d2v",IDCT=2,Ipp=False,CPU=6)
Lanczosresize(704,576)
Letterbox(8,8,8,8)
Trim(0,122970)
ConvertToYUY2()

Good work & test :)