digitalFAQ.com Forums [Archives] - Still having problems with CQ prediction

I've been into KVCDs for quite some time now. (Actually.. its nothing when compared to many other ppl in these forums :D )

But one thing which has still evaded me is the CQ prediction. I've tried ToK, ACP, CQMatic and the Manual Prediction from here. But nothing has given me correct CQ.

Most of the time my final MPG file will be smaller than expected. Like yesterday, I converted one of my XVID files to KVCD. I used CQMatic to predict the CQ. And when I used the CQ predicted by CQMatic, I got a mpg file which was of size ~ 700MB(after multiplexing with Audio). This is like 100MBs shorter than the desired size. Later, I got to know that CQMatic isnt accurate when it comes to AVI files. Still, when I burn it on a CD as KVCD and play it on my standalone player, both the audio and video quality was pretty good. But I guess it would have better if the size of the MPG was around 800MB.

Later, I tried to predict the CQ required for some other XVID file that I had using the manual prediction method given here. The AVS script that I used is given below.

Code:

## DLL Section ## 

# 

LoadPlugin("D:\KVCD\AviSynth\AviSynth 2.5\plugins\MPEG2Dec3.dll") 

LoadPlugin("D:\KVCD\AviSynth\AviSynth 2.5\plugins\GripFit_YV12.dll") 

LoadPlugin("D:\KVCD\AviSynth\AviSynth 2.5\plugins\STMedianFilter.dll") 

LoadPlugin("D:\KVCD\AviSynth\AviSynth 2.5\plugins\asharp.dll") 

LoadPlugin("D:\KVCD\AviSynth\AviSynth 2.5\plugins\UnFilter.dll") 

LoadPlugin("D:\KVCD\AviSynth\AviSynth 2.5\plugins\UnDot.dll")



#You might need to change this

Import("D:\KVCD\AviSynth\AviSynth 2.5\Functions\functions.avs") 



#Variables to be set

SampleSize=0

CQ=0

movie="Sample Movie"  

cds=1 

cdtime=80 

MinAudioBitrate=112 

factor=0.98 





# 

#### 

## Main section and static filters ### 

# 

AviSource("Q:\Sample Movie.avi",false)

# 

undot() 

Limiter() 

asharp(1, 4) 

GripCrop(352, 240, overscan=2, source_anamorphic=false)

GripSize(resizer="BicubicResize") 

STMedianFilter(3, 3, 1, 1 ) 

MergeChroma(blur(1.5)) 

MergeLuma(blur(0.1))

# 

# 





## Linear Motion Adaptive Filtering ## 

# 

# ( Portions from AviSynth's manual ) 

# This will apply variable temporalsoften 

# and variable blur. 

# Both filters are active at all times, and work inversely proportional to the 

# activity, measured from current frame to next frame. 



ScriptClip(" nf = YDifferenceToNext()" +chr(13)+ "unfilter( -(fmin(round(nf)*2, 100)), -(fmin(round(nf)*2, 100)) ).TemporalSoften( fmin( round(2/nf), 6), round(1/nf) , round(3/nf) , 1, 1)  ") 



# 

# 

# 



GripBorders() 

#LetterBox( Your_Values_Here ) # Depends on situation. Use MovieStacker! 

Limiter() 





prediction(movie, cds , cdtime, factor, SampleSize, CQ, MinAudioBitrate). Sampler(length=24)

# 

# 

## Functions ### 



function fmin( int f1, int f2) { 

  return ( f1<f2 ) ? f1 : f2 

} 

  

# 

####

The file functions.avs is given below

Code:

#------------------------------# 

# Prediction Function # 

#------------------------------# 



function Prediction (clip c, string "movie", int "cds", int "cdtime", float "factor", float "SampleSize", float "CQ", int "MinAudioBitrate") 



{ 

movie = default(movie,"No_Title_Defined") 

factor = default(factor,0.98) 

cds = default(cds,0) 

cdtime = default(cdtime,80) 

SampleSize = default(SampleSize,0) 

CQ = default(CQ,0) 

MinAudioBitrate = default(MinAudioBitrate,128) # lowest audio bitrate wanted 



MovieLength = framecount(c)/framerate(c) 

FramesTotal = framecount(c) 

FramesSample = framecount(Sampler(c,length=24)) 

CD_Size = (((cdtime*10)-5)*1024)*cds 



MinAudio = (((((CD_Size-((MinAudioBitrate/8)*Movielength))*FramesSample)/FramesTotal)*factor))/1000 

EncodedSample = (SampleSize==0)? false: true #-------------------------------------------------------------------- 

SampleSize = (SampleSize==0)? 1: SampleSize # if sample size =0 mb then no previous sample. 

CQ = (CQ<=0)? 1 :CQ # set cq between 1-100 

CQ = (CQ>100)?100 : CQ #--------------------------------------------------------------------- 

SampleSizeMessage=(EncodedSample==false)?"no sample encoded yet **********": "Sample Size (mb) : " 

SampleCQMessage=(EncodedSample==false)?"no sample encoded yet **********":"Sample CQ : " 

TryCQValueMessage=(EncodedSample==false)?"no sample encoded yet ******":"Try CQ value of : " 



d=BlankClip(c,1,544,480). 



\Subtitle(movie, 272,40, halo_color=$5500FF, text_color=$FFFFFF, font="Courier", size=27, align=5). 



\Subtitle("Total Frames : " + String(FramesTotal), 80,60). 

\Subtitle("Sample Frames : " + String(FramesSample), 80,80). 

\Subtitle("Time (sec) : " + String(round(Movielength)), 80,100). 

\Subtitle("Time (min) : " + String(round(Movielength/60)), 80,120). 



\Subtitle(SampleSizeMessage+LeftStr(String(SampleSize),5), 270,60,text_color=$FF0000). 

\Subtitle(SampleCQMessage + LeftStr(String(CQ),4),270,80,text_color=$FF0000). 

\Subtitle("Min. Audio Bitrate : "+ String(MinAudioBitrate)+" kbps",270,100,text_color=$FF0000). 

\Subtitle("Desired Sample Size (mb) : " + LeftStr(String(MinAudio),5),270,120,text_color=$FF0000). 

\Subtitle(TryCQValueMessage + LeftStr(String((MinAudio*CQ)/SampleSize),4),270,140,text_color=$FF0000). 



\Subtitle("Desired Sample Size (mb) for " + String(cds) + " CD(s):", 80,165, text_color=$FFFFFF). 

\Subtitle("- audio at 96 kbps : " + LeftStr(String((((((CD_Size-((096/8)*Movielength))*FramesSample)/FramesTotal)*factor))/1000),5),80,185). 

\Subtitle("- audio at 112 kbps : " + LeftStr(String((((((CD_Size-((112/8)*Movielength))*FramesSample)/FramesTotal)*factor))/1000),5),80,205). 

\Subtitle("- audio at 128 kbps : " + LeftStr(String((((((CD_Size-((128/8)*Movielength))*FramesSample)/FramesTotal)*factor))/1000),5),80,225). 

\Subtitle("- audio at 160 kbps : " + LeftStr(String((((((CD_Size-((160/8)*Movielength))*FramesSample)/FramesTotal)*factor))/1000),5),80,245). 

\Subtitle("- audio at 192 kbps : " + LeftStr(String((((((CD_Size-((192/8)*Movielength))*FramesSample)/FramesTotal)*factor))/1000),5),80,265). 

\Subtitle("- audio at 224 kbps : " + LeftStr(String((((((CD_Size-((224/8)*Movielength))*FramesSample)/FramesTotal)*factor))/1000),5),80,285). 



\Subtitle("Average Bitrate (kbps) : "+ String(Round(((((CD_Size-((128/8)*Movielength)))/Movielength)*8)*1)), 80,328). 



\Subtitle("Predicted Audio File Sizes (mb):", 80,373, text_color=$FFFFFF). 

\Subtitle(" 96 kbps = " + String(Round(((( 96*1024)/8)*Movielength)/1048576)),80,395). 

\Subtitle("112 kbps = " + String(Round((((112*1024)/8)*Movielength)/1048576)),80,416). 

\Subtitle("128 kbps = " + String(Round((((128*1024)/8)*Movielength)/1048576)),80,437). 

\Subtitle("160 kbps = " + String(Round((((160*1024)/8)*Movielength)/1048576)),270,395). 

\Subtitle("192 kbps = " + String(Round((((192*1024)/8)*Movielength)/1048576)),270,416). 

\Subtitle("224 kbps = " + String(Round((((224*1024)/8)*Movielength)/1048576)),270,437). 



\BilinearResize(width(c), height(c)) 

return d+c+d 

}

I was able to view the avs file in WMP. Now, the first time, I set the variables SampleSize and CQ to 0 when I called the function prediction. And once encoding is done, I got the file size and CQ and substituted them to the old values. The suggested CQ (displayed as a subtitle on the first screen) was 90. I did it again and now the suggested CQ was 68. Then it started to oscillate between ~69 and ~91. The required sample size was 11.6Megs. But with CQ of 68, the sample size size was ~8Megs and with CQ of 91, the sample size was ~13-14Megs.

(Edit: When I encoded with TMPGenc, I had set the stream type as Video Only. Does it make any difference?)

So, I stopped using the suggested CQ (that was displayed as a subtitle on the screen) and made my own trial and error encodings with different CQs until I got a file size ~ equal to the required sample size. And, I used this CQ to encode my movie. I got a nice MPEG file of size 787MB.

But, the formula that kwag has suggested didnt work. What could be the reason. Could it be because the source was AVI?. What am I doing wrong.