2 pass / crf encoding ...

lariva · 17th January 2010, 01:20

If anyone knows the answer, would be much appreciated:

With 2 pass encoding we are typically setting target bitrate and use two pass run to fit

With crf setting we are going for constant quality at whatever the bitrate is.

When I encode my movies, I (as well as possibly everyone else) look for two things:

1) good quality (crf 22 for me)
2) smallest possible file i can get

Thus what I was doing was this:

first run pass with well defined crf, once the encode would complete, it would report average bitrate, that was the rate that was "needed" to achieve the quality I was looking for.

second run would be a two pass encode with a target bitrate of the bitrate reported to by the first run.

The assumption is/was is that the two pass encode would result in a better quality for a given rate and if the rate was sufficuent for a defined constant quality, net output would be 'better'/'smoother' since it would have the benefit of both, the quality and two pass analysis.

the question is: is the assumption wrong

p.s.
during both CRF encode i kept certain things (like bframes / ref frames) and ommited others for speed..

sneaker_ger · 17th January 2010, 01:30

It is generally assumed that an encode with CRF and a second pass with the same bitrate as the first pass (done with CRF) will result in practically identical results(Given the same settings for both passes were used). In short: You are wasting your time by doing a second pass. You can simply save that time or invest it in higher settings for your CRF encode.

detmek · 17th January 2010, 01:58

Agree. Using CRF for slow first pass and same average bitrate you got from it for second pass will give you maybe 1-2% better quality. At the cost of time (twice as long).
Use that only if you use preset placebo to get a litle more quality (if you have time).
Otherwise, just use slower settings and you will get more quality compared to your method.

17th January 2010, 01:20	#1 \| Link
lariva Registered User Join Date: Feb 2007 Posts: 20	2 pass / crf encoding ... If anyone knows the answer, would be much appreciated: With 2 pass encoding we are typically setting target bitrate and use two pass run to fit With crf setting we are going for constant quality at whatever the bitrate is. When I encode my movies, I (as well as possibly everyone else) look for two things: 1) good quality (crf 22 for me) 2) smallest possible file i can get Thus what I was doing was this: first run pass with well defined crf, once the encode would complete, it would report average bitrate, that was the rate that was "needed" to achieve the quality I was looking for. second run would be a two pass encode with a target bitrate of the bitrate reported to by the first run. The assumption is/was is that the two pass encode would result in a better quality for a given rate and if the rate was sufficuent for a defined constant quality, net output would be 'better'/'smoother' since it would have the benefit of both, the quality and two pass analysis. the question is: is the assumption wrong p.s. during both CRF encode i kept certain things (like bframes / ref frames) and ommited others for speed..

17th January 2010, 01:30	#2 \| Link
sneaker_ger Registered User Join Date: Dec 2002 Posts: 5,565	It is generally assumed that an encode with CRF and a second pass with the same bitrate as the first pass (done with CRF) will result in practically identical results(Given the same settings for both passes were used). In short: You are wasting your time by doing a second pass. You can simply save that time or invest it in higher settings for your CRF encode.

17th January 2010, 01:58	#3 \| Link
detmek Registered User Join Date: Aug 2009 Posts: 463	Agree. Using CRF for slow first pass and same average bitrate you got from it for second pass will give you maybe 1-2% better quality. At the cost of time (twice as long). Use that only if you use preset placebo to get a litle more quality (if you have time). Otherwise, just use slower settings and you will get more quality compared to your method.