[x265] Bitrate calculation is off by 5:1 for FullHD video

Mario *LigH* Rohkrämer contact at ligh.de
Fri Feb 21 16:01:29 CET 2014


Am 18.02.2014, 18:31 Uhr, schrieb Steve Borho <steve at borho.org>:

> On Tue, Feb 18, 2014 at 6:55 AM, Mario *LigH* Rohkrämer  
> <contact at ligh.de>wrote:
>
>> Just the last 2 examples of encoding 60 seconds from ToS in 1080p:
>>
>> a) CRF 24 => 11402140 B * 8 b/B : 60 s = 1520.3 kbps (x265 reports  
>> 7597.54
>> in CSV)
>> b) CRF 18 => 26700793 B * 8 b/B : 60 s = 3560.1 kbps (x265 reports
>> 17796.64 in CSV)
>>
>> It seems that x265 reports about 5x the bitrate of the result, but only
>> for this large video dimension.
>>
>> Another test with a small video dimension (640x272) hits the bitrate
>> rather correctly:
>>
>> c) fast: 709078 B * 8 b/B : 52.208 s = 108.6 kbps (x265 reports 107.87  
>> in
>> CSV)
>> d) ultrafast: 5325541 B * 8 b/B : 52.208 s = 816.0 kbps (x265 reports
>> 815.27 in CSV)
>>
>> CSV logs are available on request.
>
>
> I'm a bit confused by which numbers you are comparing.  Are you using
> frame-by-frame CSV or just the single-line-per-run CSV?  And are you
> comparing the file size or the theoretical size?

Sorry, missed this mail in the threads...

I'm comparing

a) the bitrate calculated by the size of the written HEVC video file  
divided by its playing time

with

b) the bitrate reported in the last line of a frame-by-frame CSV log  
(level 3), which summarizes the result of the whole job

-- 
__________

Fun and success!
Mario *LigH* Rohkrämer
mailto:contact at ligh.de
 


More information about the x265-devel mailing list