Log in

View Full Version : Fastest settings for the x264 codec?


gobiaallloo
2nd June 2006, 09:26
Hi folks!

I know i am a complete newbie for the x264 codec but as far as i know encoding using x264 codec takes a lot of time but saves disk space... but a friend of mine records some tv shows and edits them(i.e remove ads) and encodes them using the x264 codec in just 11 minutes! He uses VirtualDub MPEG2 for editing but but is it possible to encode a 21 min show in just 11minutes?

I have tried a lot to know the cause behind it but i was unsuccessful... pls tell me the fastest settings for encoding a mpg1 file using the x264 codec...

Waiting for replies,

Regards,
Gaurav(gobiaalloo)

GodofaGap
2nd June 2006, 09:34
Do you and your friend have the same type of machine?

Fastest setting would probably be:

--subme 1 --me dia --analyse "none"
On normal setting using b-frames is usually faster for me, but I don't know with these. Of course you can specify --threads X to suit your processor.

gobiaallloo
2nd June 2006, 09:36
thanks for ur reply but can u make it a bit simpler i m a complete n00b to x264 codec

GodofaGap
2nd June 2006, 09:40
If you want a simpler answer, you will have to refine your question. You asked for the fastest settings and I gave them. I have no idea what programs you are trying to use, and I have no idea what level of expertise "complete n00b" refers too.

gobiaallloo
2nd June 2006, 10:07
i m trying to use VirtualDub MPEG2 for encoding and editing the video..

DarkZell666
2nd June 2006, 12:27
lol, you want the VFW settings then (I don't have x264 installed i'm at work, so someone else will have to give them :p)

As for "is it possible to encode 21min show in 11min" : yes, if you resize to something like 512*384 and use the fastest settings :)

BUT !! Using the fastest settings will generally take up more space or give "not too good" quality.

Kostarum Rex Persia
2nd June 2006, 15:35
Using VirtualDub MPEG2 is easy, just set resize filter in Filters meny to 512x384 or less, and use --subme 3 --me dia.

Using --subme 1 is very bad, because quality is bad.

imcold
2nd June 2006, 15:49
Disable all partitions, set partition decision to 1, method to diamond search, disable chroma ME, CABAC, trellis, deblocking filter and you will get a huge speed boost :D

gobiaallloo
2nd June 2006, 16:25
can any1 tell me how do i use "--subme 3 --me dia." command?? I am Using Vdub MPEG2 which uses VfW codec..

shon3i
2nd June 2006, 22:01
Go to more tab

Partition Decision - 3
Metod - Diamond Search

Sagittaire
2nd June 2006, 22:24
fastest setting are
--subme 1 --me dia --analyse "none" --no-chroma-me

Using --subme 1 is very bad, because quality is bad.

done always better quality than XviD for same size ... :devil:

JarrettH
4th June 2006, 07:37
This place is so user friendly :p

Livesms
9th June 2006, 07:22
fastest setting are
--subme 1 --me dia --analyse "none" --no-chroma-me

done always better quality than XviD for same size ... :devil:

Please tell us your Xvid setting for that case :)

Sagittaire
9th June 2006, 11:00
Please tell us your Xvid setting for that case :)

With all setting that you want ... ;)

But you have certainely magical XviD setting with magical CQM ... like all XviD user's ... lol

Livesms
9th June 2006, 12:07
With all setting that you want ... ;)

But you have certainely magical XviD setting with magical CQM ... like all XviD user's ... lol

Can you explain.

You said that x264 with "--subme 1 --me dia --analyse "none" --no-chroma-me" is always better quality than XviD for same size ...

I asked you to explain what xvid settings do you mean.

Look's like Xvid with Qpel, BVops, ME 6, VHQ 4, Chroma ME, VHQ for BVOPs will be not better then x264 with ME 1fastest and Diamonds search (no partition search, no b-frame, no Chroma me and so).

Or you are kidding?

Sharktooth
9th June 2006, 13:49
Always means ALWAYS... dont you think?
with -no-cabac it will be faster though... but will loose a lot of compression

Sagittaire
9th June 2006, 14:54
Can you explain.

You said that x264 with "--subme 1 --me dia --analyse "none" --no-chroma-me" is always better quality than XviD for same size ...

I asked you to explain what xvid settings do you mean.

Look's like Xvid with Qpel, BVops, ME 6, VHQ 4, Chroma ME, VHQ for BVOPs will be not better then x264 with ME 1fastest and Diamonds search (no partition search, no b-frame, no Chroma me and so).

Or you are kidding?

kidding, me, no ... lol

x264 fastest setting is better than xvid slowest setting ... you want a little demonstration ... ???

DeathTheSheep
9th June 2006, 18:00
Yeah, I know it's unbelievable at first :P
I got similar results to Sagittaire-- AVC truly is "the future of video encoding!"
Though I've never actually tried --subme 1... 3's about the lowest I decided on... tch.

AlexI
11th June 2006, 08:13
Disable all partitions, set partition decision to 1, method to diamond search, disable chroma ME, CABAC, trellis, deblocking filter and you will get a huge speed boost :D

in terms of bang-for-the-buck, disabling cabac and deblocking is not worth it... yes it is faster, but there are other things which will give a similar speedup at better quality.

Probably the fastest "useful" settings are --me=dia --subme=2 --analyse=none --bframes=1 --no-b-adapt --no-chroma-me. Beyond that you're better off simply dropping the resolution ;)

Sharktooth
11th June 2006, 13:48
you can always reduce ME Range to 12 or 8...

Livesms
11th June 2006, 13:54
you can always reduce ME Range to 12 or 8...

And how it will change quaility?

foxyshadis
11th June 2006, 17:30
Not that much, even at low bitrates, and less as it gets higher. There was a test a while back comparing 8/16/32 and if you need speed 8 doesn't seriously impact quality except in frentic scenes.

imcold
11th June 2006, 23:29
i wonder why is that: is it because the cost of storing intra-block (smaller range = fewer matches = more i-blocks) lower than cost of motion vector, which is far (32px) from prediction?

akupenguin
12th June 2006, 09:05
Why the quality impact of reducing search range is small? Simply because the vast majority of all blocks have the correct final motion vector within a few pixels of the search's starting point.
Before it even gets to the part of motion estimation controlled by me_method and me_range, x264 first tries up to 10 candidate motion vectors from various prediction methods, and keeps the best one. Chances are, at least one of the 10 predictions is close to the real mv.