1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

CQMatic for Tmpgenc

Discussion in 'MPEG-1 and MPEG-2 encoding (AVI to DVD)' started by shiroh, Aug 10, 2004.

  1. shiroh

    shiroh Guest

    this is a tool to calculate or predict the correct CQ for a targeted filesize.

    basically it takes a some part/frames of the clips and run it through 10 cycles with different quant and you'll get the final quant.

    my test source is a 22 minute anime 30500 frames at 24fps.
    the 10 cycles took 40 minutes to complete and the video encode also took 40 minutes, both at motion estimate precision.
    while at 2 pass it took 1 hour 30 minutes to fully finish.
    a marginally 10 minute faster.

    the final result is impressive. the final file is exactly what i wanted :)


    why you want to use it ?

    1) if you don't trust tmpgenc 2 pass rate control. CQ is better than 2 pass, at least i think so. with 2 pass the encoder needs to decide what to give based on the programmers algorithm. if it doesn't mess up.
    but with CQ it uses a constant Quantizer (the Q in CQ for what i care is Quantizer) which hard to mess up. ( for more information do a search about quantization)
    to put it simply, CQ is more precise in than 2 pass decision. yet again, i think it is.

    2)if you want to encode the video using higest quality motion precision. i'm not sure how this works actually, but in 2 pass, the motion search will be use for whatever you set up, if you chose highest that will increase the time greatly.
    what i have in mind here you can use CQmatic to calculate the CQ using a faster mode, and change to the desireable mode when the real encoding takes place. since CQmatic only predicts the CQ so the video quality is not affected by the change of modes, i think. :) but there are consequences.
    i've did a simple test with 1000 frames of the same source.
    motion estimate yields 9,748KB and with highest quality 9,478KB.
    so if you do change the to a higher setting later, the file will get smaller. bah, i'm not good at explaining, think about it.

    3)if you want a faster encode, yeah 10 minutes faster is something, bt i have something in mind. like i do with a serial episode anime, it tends to have the similar style/traits of motion, lighting, texture all that sort of stuff.
    which means the CQ calculated earlier might result in a similar filesize. ecspecially tv sitcoms, with its static background and similar backdrops. just don't be to accurately perfect like you want a 700MB perfect. just lower down your video filesze target. you can always up the audio. i do audio seperately so its not a problem. and don't forget to cross your fingers ;)


    so this is one great tool you should look out for. at

    http://kvcd.net/forum/viewtopic.php?t=5145

    credits and thanks to the developer, kwag, of kvcd.net.

     
  2. shiroh

    shiroh Guest

    at last i found some time to watch and compare results from using CQ and 2 pass VBR.

    both are at same size, almost.
    and CQ turn out to be better, compared on TV. with CQ there were less dancing blocks, ecspecially during motion and fades.

    i would say if you want to use 2 pass, its better if its on a high average bitrate, like say....3000-4000 average. but thats just me.

    the average of my test were at 2050kbps. for a sometime dark and sometimes bright anime.

    so CQ is the way to go, as 2 pass decision looks crap compared, at least on tmpg 2.5 plus. i wonder if they work it out on the 3.0
     
  3. Minion

    Minion Senior member

    Joined:
    Sep 28, 2003
    Messages:
    5,623
    Likes Received:
    0
    Trophy Points:
    116
    It seems Like a Lot of Time (40 Minutes) just to calculate the File size?!?
    If you want Faster encodeing why not just use a Faster Encoder??
    Something like CCE SP 2.67 is up to 5 Times as fast as Tmpgenc and Produces better quality Mpeg2 files at a Lower Bitrate....

    Just a Thought??
     

Share This Page