View Full Version : Constant Quality encoding h.264
atreides93
17th July 2008, 02:14
I'm really impressed with constant quality encoding. I tried it today with a setting of 26 and the resulting file was only 2 gigs for a 2 hour movie. I thought for sure it would look horrible. Strangely enough, it looked great, and that was 1920x1080. Its ironic because earlier I did a 2 pass encode and got it to fit in 5 gigs thinking how great that was.
The only problem, is the aspect ratio is messed up, the movie is squashed...I'll have to investigate that.
Dark Shikari
17th July 2008, 02:14
For aspect ratio, --sar
wyti
17th July 2008, 02:20
Yes constant Quality mode is great, you should always use that mode, if you don't care about the final size of the file.
Ranguvar
17th July 2008, 02:34
The only problem, is the aspect ratio is messed up, the movie is squashed...I'll have to investigate that.
You can change the aspect ratio in the container, so no need to re-encode :)
atreides93
17th July 2008, 03:02
I'm trying that now with h264info.
It blows my mind how many excellent free utilities are out there for this stuff.
Dark Shikari
17th July 2008, 03:05
I'm trying that now with h264info.
It blows my mind how many excellent free utilities are out there for this stuff.h264info changes it in the stream--in the container, its even easier. MKVmerge for MKV, YAMB for MP4.
audyovydeo
17th July 2008, 08:30
I'm trying that now with h264info.
It blows my mind how many excellent free utilities are out there for this stuff.
mp4box yourfile.mp4 -par [trackID]=[AR]
mp4box myfile.mp4 -par 1=32:27
cheers
a/v
Comatose
17th July 2008, 12:17
Hm, I have a question: is CRF mode as efficient as bitrate mode? that is - if I let CRF mode encode at 20, and then encoded the same video with 3 passes to the filesize I got from CRF mode, would the quality be identical?
foxyshadis
17th July 2008, 12:30
It would be close enough that you probably wouldn't be able to tell the difference even with frame-by-frame comparison. There might be differences at high quantizer, but you'd be very hard-pressed to decide which is actually better.
Nikos
17th July 2008, 13:06
With constant quality encoding (CRF) the suggested setting is
--aq-mode 2, move bits between frames (default)
or
--aq-mode 1, avoid moving bits between frames ?
Dark Shikari
17th July 2008, 14:50
With constant quality encoding (CRF) the suggested setting is
--aq-mode 2, move bits between frames (default)
or
--aq-mode 1, avoid moving bits between frames ?The former. Don't touch that unless you're doing CBR or something.
CRF is basically equivalent to (potentially slightly better than) twopass bitrate at the same bitrate, quality-wise.
Comatose
17th July 2008, 17:31
It's great to hear that :p I thought it might not try to optimize the size as much since it cares less about size and more about quality :E
Sharktooth
17th July 2008, 17:58
CRF = Constant Rate Factor... constant qulity is a consequence...
type1encryption
17th July 2008, 20:25
CRF is basically equivalent to (potentially slightly better than) twopass bitrate at the same bitrate, quality-wise.
so...why is that?
Dark Shikari
17th July 2008, 20:40
so...why is that?Because the purpose of twopass bitrate is to pick the correct ratefactor that would give you the resulting bitrate ;)
A constant ratefactor is the ideal--bitrate mode always has to adapt the ratefactor to reach the target, even on the second pass.
Of course, bitrate mode benefits from a couple things in twopass that help it also, which is why CRF isn't universally better. On average, they're roughly the same.
type1encryption
17th July 2008, 21:07
but aren't these decisions made more efficiently in 2pass, because of b-frames etc..
or is crf constantly 2-passing it internally up to the next keyframe?
Dark Shikari
17th July 2008, 21:17
but aren't these decisions made more efficiently in 2pass, because of b-frames etc..B-frame decision is only made on the firstpass anyways.or is crf constantly 2-passing it internally up to the next keyframe?Not necessary.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.