VideoHelp Forum




+ Reply to Thread
Results 1 to 8 of 8
  1. Member
    Join Date
    Dec 2006
    Location
    United States
    Search Comp PM
    I was hoping for some help and some critiques in trying to get the highest quality xvid avi from high definition digital cable captures. I have seen the type of quality you can get from this type of conversion and was hoping to reach that level and beyond. My goals are to have full 1080p avi with 5.1 sound. So far I have been getting pretty good results with the average size file of 1,400 for a 45 min show.

    My method is 1st I run my .ts through mpeg2repair
    2nd demux with DGIndex
    Then I run this avisynth script through VDM:

    [code]
    Tfm()
    Tdecimate()
    RemoveGrain(mode=1)
    RemoveGrain(0,11)

    Then I run a 2 pass encode with the video and audio demuxed from DGIndex. I set the target size according to a ratio, basically 1 min = roughly 38 mbs.

    Some of the specific questions I have are:
    How does my avisynth script look? Any advice one way or the other?

    What are the precise settings in VDM to achieve the high quality encodes that I am trying to achieve. I have read through 3 guides about what each setting does, but not any advice on what the settings should be, and how it affects overall quality. Things like should I use BVOPs? I think that I understand how they are used. Basically they take information from the frame ahead and the frame behind and compresses them, right? So those 2 frames that this info comes from, does that get dropped?

    I’ve read about the quantization and different quantization matrices; one for high bitrate, low bitrate, ultra low, and so on. Is there one that any uses to encode high quality hdtv captures to xvid.

    How about quantizer restrictions? I’ve read for min that 2 is better than 1, because 1 results in big file size with not much more quality than 2. So isn’t max 31 too high for what I am trying to do?

    Is it a good idea to use VHQ mode 4? I am not worried about encoding times at all, I’ve got time, as long as we’re not talking about days and days.

    I am not even sure if tweaking these settings will make that much of a difference. Will it?

    I’ve had some issues where my encodes get a bit jumpy / blocky (hard to describe) in high motion and pan and scan scenes. Like something is flashing in between frames horizontally. There are no artifacts on the frames when I look at them one by one, but when played in succession, I guess it isn’t as smooth and nice as when there isn’t any motion or little motion. Any ideas to remedy this?

    Here are 2 links to 10 secs of my encodes thus far if anyone is interested.

    http://www.filecrunch.com/file/~7ed61y
    Right before the first guy swings you’ll see what I am talking about in terms of high motion scene problems that I am experiencing, and then other guy retaliates.

    http://www.filecrunch.com/file/~tlgoo6
    And again as the last building is in a panning shot you’ll see it again.

    I know that quality is a subjective term, so I guess I can put it this way. If I am willing to set aside up to 1,400 mbs per file, what settings should I use to get really great visually appealing results?


    Any advice or information would be so appreciated.
    This forum rocks. Thanks.

    I do apologize for the length of this post. Sorry.

    Happy holidays.
    Quote Quote  
  2. Member edDV's Avatar
    Join Date
    Mar 2004
    Location
    Northern California, USA
    Search Comp PM
    Without getting into details, I'd offer some strategic input.

    720p/59.94 to 1080p/59.94 is very doable but any 1080p TV will do it's own upscale and other sets won't accept 1080p/59.94 or 1080p/23.976 directly. Your xvid player may need to convert it back to 1080i or 720p. You will get better compression from 720p than from upscaling and then compressing. 720p/59.94 movies can be stored as 720p/23.976 with proper frame decimation.

    Similar issues with 1080i but interlace xvid may be currently impossible. Within another year VC-1 will mature to preserve native 1080i so the 1080p and other HDTVs can do their own conversion to progressive. Movie 1080i should be IVTC'd to 1080p/23.976. I agree with that.

    I guess what I'm suggesting is other than removal of repeating film fields and frames, it may not need fixing, it just needs compression.
    Quote Quote  
  3. Since you're choice of Xvid at 1400 kbps is rather arbitrary, why don't you use single pass target quantizer mode. Pick the quality you want and encode in one pass. It's faster and gives you a known quality, as opposed to 2-pass VBR which gives you a known size but unknown quality.

    As far as I can tell with Xvid, B frames only save bitrate because they are encoded at a lower quality (higher quantizer). I don't use them.

    I like to set the Max I-Frame Interval to 100. It makes seeking easier and faster. You might want to go even lower with HD encodes.

    VHQ Mode above 1 - Mode Decision are much slower and don't give much more quality, or with constant quantizer, not much smaller files. Usually not worth it.
    Quote Quote  
  4. Member
    Join Date
    Dec 2006
    Location
    United States
    Search Comp PM
    Originally Posted by jagabo
    Since you're choice of Xvid at 1400 kbps is rather arbitrary, why don't you use single pass target quantizer mode. Pick the quality you want and encode in one pass. It's faster and gives you a known quality, as opposed to 2-pass VBR which gives you a known size but unknown quality.
    That would give me file sizes much larger than 1,400 mb at the quality I want without “smart” compression.

    I’ve tried diff target quantizer settings. 4 gives me good quality, too big a size. Anything > than 4, picture not so great, file size getting smaller. Needless to say, anything < than 4, file size gets bigger.

    How can you tell if a file is encoded with B frames in it?
    Quote Quote  
  5. Member
    Join Date
    Dec 2006
    Location
    United States
    Search Comp PM
    Originally Posted by jagabo
    As far as I can tell with Xvid, B frames only save bitrate because they are encoded at a lower quality (higher quantizer). I don't use them.
    After some experimenting, B frames at the setting max consecutive 2 / 1.5 / 1 saves quite a bit of bitrate: 8000 vs. 4000 kbps
    Quote Quote  
  6. Banned
    Join Date
    Oct 2004
    Location
    Freedonia
    Search Comp PM
    Is there a good reason you have decided on 1080p, because nothing is broadcast in that. If your source is 720p, I advise sticking with that. I have gotten truly excellent results in converting 1080i to 720p Dixv/Xvid (I'm not a religious zealot that insists on one or the other) at bit rates of 3000 to 3500. I am guessing that you are NOT encoding movies, so edDV's suggestion that 1080i should be ITVC'd may not apply to you.

    I messed with VC-1 and H.264 a little. Couldn't get VC-1 to work on my sources. H.264 worked, but it takes FOREVER and the documentation is almost non-existent on what the various encoding options do. Divx/Xvid works for me right now. I suggest this not to prove you are wrong, as if I could, but merely to point out a very easy alternative that I am getting excellent results from.
    Quote Quote  
  7. Originally Posted by drj
    Originally Posted by jagabo
    Since you're choice of Xvid at 1400 kbps is rather arbitrary, why don't you use single pass target quantizer mode. Pick the quality you want and encode in one pass. It's faster and gives you a known quality, as opposed to 2-pass VBR which gives you a known size but unknown quality.
    That would give me file sizes much larger than 1,400 mb at the quality I want without “smart” compression.
    In my experience, if you create files of the same size (same average bitrate) with single pass constant quantizer encoding and 2-pass VBR encoding they will look almost identical.

    Originally Posted by drj
    After some experimenting, B frames at the setting max consecutive 2 / 1.5 / 1 saves quite a bit of bitrate: 8000 vs. 4000 kbps
    That also differs from my experience. I usually see only a 25 percent reduction in average bitrate between no B frames and Xvid's normal 2/1.5/1 setting. And again, this is only because the B frames are saved with a higher quantizer. If you force Xvid to use the same quantizer for B frames the files come out the same size.

    I usually encode at CQ 2 or 3.
    Quote Quote  
  8. Member edDV's Avatar
    Join Date
    Mar 2004
    Location
    Northern California, USA
    Search Comp PM
    If compression is the first goal and you can live with telecined 1080i reduced in half to 29.97 fields (ewww gag me) then you guys are missing the joys of 960x540p the refuge of early HDTV processors. 960x540 is single field 1080i with horizontal resolution reduced in half. Result is 1/4 bitrate before further compression.

    One forth 1920 is 480x540p. One third horizontal would be 640x540p. Integer math like 1/2x 1/4x results in less softening.
    Quote Quote  



Similar Threads

Visit our sponsor! Try DVDFab and backup Blu-rays!