VideoHelp Forum
+ Reply to Thread
Page 1 of 2
1 2 LastLast
Results 1 to 30 of 48
Thread
  1. Member
    Join Date: Sep 2004
    Location: australia
    Search Comp PM
    hi there, taking a 1080/50p m2ts file and outputting it to 720/50p m2ts or mp4, is it better to do the encode with 2 passes, or is one pass ok.

    just reading conflicting stories about this, just like reading different opinions about whether you should do a single or double pass encode when converting HD video (h264) to Mpeg2 (dvd) format, in which this case i always do a 2 pass encode anyway.

    cheers
    Quote Quote  
  2. There is no such thing as 1 pass (CBR, CRF, CQ) or 2pass (VBR) is better or worse , encoder always does its best ,

    1. The question is I have video of THIS length and I want to squeeze it to THAT size. Using 2pass VBR.

    2. Or I have video that I care to compress and I want THIS quality, using 1pass CRF.

    It is up to you what approach you choose. #2 is flexible though, you can try to encode just 10 seconds and you know what quality it is giving out and then you change CRF accordingly . After that you are set for many, many projects of yours. As for 2 pass VBR, you have to encode it all, and then you go scene after scene to check what you got. But as I said in case of restricted space this is what you have to do to be most effective. And you never know what you are going to get, using 2pass VBR.

    Do not use 1 pass VBR. 1pass CBR also , not much effective, , except DVD in this case, if you know that DVD is not longer than say hour and 10 minutes, you can use it, waste of space , sure, but who cares how full the DVD is.
    Some encode AVCHD 28 Mbps or so (not sure about exact bitrate) , there is different bitrates available I guess, that is 1pass CBR, well, whatever works, ...
    Last edited by _Al_; 9th Apr 2014 at 18:52.
    Quote Quote  
  3. Member
    Join Date: Sep 2004
    Location: australia
    Search Comp PM
    hi, im not trying to convert to any set size, i use VRD and it has built in quality presets that we can use, or we can modify those presets to suit our own needs, and it uses variable bitrates.

    the built in preset will output a 1080/50p m2ts to 720/50p m2ts @ 12Mbps and if i convert that to 720/25p (half the framerate) the output file is around 6Mbps, but i have always been using 1 pass, and have been told by the devs over at VRD to use 2 passes, and it gets very confusing.

    just as some people (including the Devs at VRD) tell me to convert 1080/50p m2ts files to Mpeg2 as progressive (their h264 to Mpeg2 preset does this automatically) there are others who say to encode them as interlaced.
    Quote Quote  
  4. If that 1pass is VBR, it is good to recommend not to use it.
    But if that 1pass is CRF , then why not, you can use it, I'd even prefer it.
    Or is it CBR? Those values 12Mbit or 6Mbit are average or just constant bitrate?

    To encode DVD from HD,50p, was discussed here too. Making 25p you loose temporal resolution. Strobing is going to be present.
    Is that software simply dropping every other frame? Then your result very is sharp but with some strobing in motion)
    Some videoeditors use blending making 25p to complicate things. Not sure what your software does.

    That is why to better use 25i, to keep that temporal resolution but, big BUT, low pass filter, or vertical blur should be used (or other method I do not know about) to not have interlace lines flickering. This vertical blur will get result not so 100 % sharp as 25p sharp resize but motion is fluid (as was recorded).

    So it depends what your software actually does with that footage, maybe that is why they recommend mpeg2 as progressive 25p knowing that is best compromise for their software. The longer shutter you shoot with, the better result with 25p, but anyway, never that smooth as original.
    Quote Quote  
  5. Originally Posted by glenpinn View Post
    just as some people (including the Devs at VRD) tell me to convert 1080/50p m2ts files to Mpeg2 as progressive (their h264 to Mpeg2 preset does this automatically) there are others who say to encode them as interlaced.
    It depends on the source whether you keep it interlaced or make it progressive. By MPEG-2 do you mean DVD? If so, progressive sources can be encoded for PAL DVD as either interlaced or progressive. Retail PAL DVDs of progressive sources (like movies) are usually encoded as interlaced. Of course, interlaced sources should always be encoded as interlaced. Do you notice there's a difference between what the source or content is like and how it's encoded? You may not be understanding what you're being told and are garbling it in the 'translation'.
    ...im not trying to convert to any set size
    Then you should use 1-pass CRF. Not 1-pass ABR or CBR. Get the quality you want and forget about the size.
    Originally Posted by _Al_ View Post
    Making 25p you loose temporal resolution.
    If the source is a progressive movie then every other frame is a duplicate and you lose nothing by making it 25fps. Or are these home movies? A sample might help.
    Last edited by manono; 9th Apr 2014 at 20:06.
    Quote Quote  
  6. Originally Posted by manono View Post
    If the source is a progressive movie then every other frame is a duplicate and you lose nothing by making it 25fps. Or are these home movies? A sample might help.
    There was mentioning 50p at the beginning of this thread, if it was about movies I would perhaps not responded at all, I thing it is 50p camcorder footage.
    Quote Quote  
  7. My mistake then. I apologize, and thanks.
    Quote Quote  
  8. Member
    Join Date: Sep 2004
    Location: australia
    Search Comp PM
    ok, this is 1080/50p avchd video from my camera, sorry if it was confusing.

    VRD converts h264 to h264 using variable bitrate, and the Devs told me to do a 2 pass when re-converting these 1080/50p files to say 720/50p or 720/25p h265 (m2ts or mp4)

    if i convert the 1080/50p file to a Pal dvd compliant Mpeg2 file i use VRD uses variable bitrate, and i have been advised by the Devs at VRD to do a 2 pass on this as well.

    i will upload some sample 1080/50p m2ts from mt camera as well as my 720/50p m2ts and 720/25p conversions using single and double pass so you can download them and see for yourself.
    Quote Quote  
  9. As for mp4 content,
    You want to know why 2pass or 1pass but you do not say what kind of 1pass you have available, CBR, VBR or CRF.
    1pass CBR is not economic at all and high bitrates are being cut off. If bitrate is high enough - some use this method for whatever reasons - 1pass only, or simply software does not have CRF or they just do not care, thinking AVCHD (1pass CBR) is more compatible, etc..
    1pass VBR is trying to stick close to bitrate value, but it is not economic also.
    1 pass CRF must have that CRF value selected, not bitrate, not sure if VRD can do it as well. 1pass VBR is not 1 pass CRF.

    DVD approach I explained too. I doubt that there is any vertical blur involved in making 25i DVD from 50p HD original. So that is why they recommend 25p out 50p original perhaps.
    Quote Quote  
  10. Member
    Join Date: Sep 2004
    Location: australia
    Search Comp PM
    VRD uses Variable bitrates for all its outputs, and if i ever convert my 1080/50p h264 files to 1080/25p, or 720/50p or 720/25p h264 then i always use their built in "Quality" mode preset (as you can see in the screenshots below)

    this is my 1080/50p m2ts to 720/50p m2ts profile, and my 24Mbps 1080/50p m2ts source file will output to 720/50p m2ts @ around 11Mbps
    http://www.mediafire.com/view/p2ip8z5rm8d9oel/VRD_720-50p_m2ts_profile.jpg

    this is my 1080/50p to 720/25p m2ts output profile, and because the framerate goes from 50p to 25p the output 720/25p m2ts file will be around 5.5Mbps
    http://www.mediafire.com/view/lod38b47obm4vn6/VRD_720-25p_m2ts_profile.jpg

    if i convert from 1080/50p to Pal dvd Mpeg2 then i use this profile, except i use the "Bitrate" mode set to 8000kb, and Mpeg2 outputs also uses a variable bitrate, not constant.
    http://www.mediafire.com/view/e46k8sny38370sb/VRD_720x576_mpeg2_profile.jpg
    Last edited by glenpinn; 9th Apr 2014 at 20:58.
    Quote Quote  
  11. Looking at those pictures, I do not know what that "Quality Mode" in videoredo does exactly. They have their own approach for encoding settings using ffmpeg, "intelligent recode" suppose to output similar quality like input, that percentage setting suppose to manipulate outcome a bit, so maybe it is CRF mode, but how it mirrors quantizer settings exactly I do not know ...

    VideoRedo uses ffmpeg so they can use quantizers for encoding now

    Inteligent recode could mean CRF 19 (or in that proximity) and percentage could mean logical increments for CRF or other settings to adjust size for outcome, just guess.
    Last edited by _Al_; 9th Apr 2014 at 21:35.
    Quote Quote  
  12. Member
    Join Date: Sep 2007
    Location: Canada
    Search Comp PM
    Originally Posted by _Al_ View Post
    Looking at those pictures, I do not know what that "Quality Mode" in videoredo does exactly. They have their own approach for encoding settings using ffmpeg, "intelligent recode" suppose to output similar quality like input, that percentage setting suppose to manipulate outcome a bit, so maybe it is CRF mode, but how it mirrors quantizer settings exactly I do not know ...

    VideoRedo uses ffmpeg so they can use quantizers for encoding now

    Inteligent recode could mean CRF 19 (or in that proximity) and percentage could mean logical increments for CRF or other settings to adjust size for outcome, just guess.
    Unless something has changed, videoredo uses mainconcept for h.264 encoding

    ffmpeg would mean libx264, which would actually be preferrable
    Quote Quote  
  13. Member
    Join Date: Sep 2004
    Location: australia
    Search Comp PM
    this is a MediaInfo screenshot of a 1080/50p m2ts converted to 720/50p m2ts using the VRD Quality mode preset.

    http://www.mediafire.com/view/qrs576w9aeov0mw/1080-50p_to_720-50p_MediaInfo.jpg
    Quote Quote  
  14. I got that from videohelp page for videoredo, I search ffmpeg and I got ffmpeg mpeg2 encoder , now I search mainconcept now and it gave me result as well, so they might use mainconcept for that quality mode ?
    MainConcept has constant quality if I remember but does it use something like constant rate factor as well like x264 so quantizer is changing a bit while there is motion?
    Anyway, that quality mode, is it constant quantizer for the purpose of this thread?
    Quote Quote  
  15. Member
    Join Date: Sep 2004
    Location: australia
    Search Comp PM
    Originally Posted by poisondeathray View Post
    Unless something has changed, videoredo uses mainconcept for h.264 encoding

    ffmpeg would mean libx264, which would actually be preferrable
    VRD tell me they use the Mainconcept encoder for Mpeg2 but not sure about h264

    i am so confused

    i am currently doing a bunch of sample outputs using a 1080/50p m2ts source file, will put the links up when they are done, and you can download them and take a look.

    - output to 1080/25p m2ts (Quality mode)
    - output to 720/50p m2ts (Quality mode)
    - output to 720/25p m2ts (Quality mode)
    - output to Pal dvd Mpeg2 (progressive & interlaced) using TMPGEnc 8000kb vbr
    - output to Pal Mpeg2 (progressive) using VRD 8000kb vbr (vrd wont let me interlace progressive source files for Mpeg2)
    Last edited by glenpinn; 9th Apr 2014 at 21:57.
    Quote Quote  
  16. Member
    Join Date: Sep 2007
    Location: Canada
    Search Comp PM
    Originally Posted by _Al_ View Post
    I got that from videohelp page for videoredo, I search ffmpeg and I got ffmpeg mpeg2 encoder , now I search mainconcept now and it gave me result as well, so they might use mainconcept for that quality mode ?
    MainConcept has constant quality if I remember but does it use something like constant rate factor as well like x264 so quantizer is changing a bit while there is motion?
    Anyway, that quality mode, is it constant quantizer for the purpose of this thread?
    mainconcept sdk engine has both constant quality (quantizer mode) and constant quality (average quantizer, similar to x264 crf) modes available, but not options are exposed in their licensed products . Having more options available means the license fee increases




    VRD tell me they use the Mainconcept encoder for Mpeg2 but not sure about h264
    Both - that's why some consumers are slightly dissatisfied with the h264 quality . It's fine if you use higher bitrates, but low to medium bitrates will expose issues more clearly
    Quote Quote  
  17. Originally Posted by poisondeathray View Post
    but not options are exposed in their licensed products . Having more options available means the license fee increases
    How they can control "quality" (Intelligent Recode" and Quality Mode" setting) not having constant quality modes available in videoredo?
    Quote Quote  
  18. Member
    Join Date: Sep 2004
    Location: australia
    Search Comp PM
    Originally Posted by poisondeathray View Post
    Both - that's why some consumers are slightly dissatisfied with the h264 quality . It's fine if you use higher bitrates, but low to medium bitrates will expose issues more clearly
    but you would not use a low to medium output bitrate if it is not deemed to be enough for that given output format.

    VRD uses "Intelligent Recode" which outputs the encoded file at what is deemed to be the right bitrate for the output you set.

    so, as an example, if i output a 1080/50p m2ts file to 720/50p m2ts, VRD encodes it at about 45% of the source files bitrate, thats how they set up their Quality Mode preset, but the output file still has a Variable bitrate according to MediaInfo, just look at my screenshot in my post up the page.

    like i already said, a 1080/50p m2ts file is 24Mbps, VRD outputs the 720/50p m2ts at 11Mbps (45% of the source bitrate) and any higher that this would be considered pointless according to the Devs.
    Quote Quote  
  19. Member
    Join Date: Sep 2007
    Location: Canada
    Search Comp PM
    Originally Posted by glenpinn View Post
    Originally Posted by poisondeathray View Post
    Both - that's why some consumers are slightly dissatisfied with the h264 quality . It's fine if you use higher bitrates, but low to medium bitrates will expose issues more clearly
    but you would not use a low to medium output bitrate if it is not deemed to be enough for that given output format.

    VRD uses "Intelligent Recode" which outputs the encoded file at what is deemed to be the right bitrate for the output you set.

    so, as an example, if i output a 1080/50p m2ts file to 720/50p m2ts, VRD encodes it at about 45% of the source files bitrate, thats how they set up their Quality Mode preset, but the output file still has a Variable bitrate according to MediaInfo, just look at my screenshot in my post up the page.

    like i already said, a 1080/50p m2ts file is 24Mbps, VRD outputs the 720/50p m2ts at 11Mbps (45% of the source bitrate) and any higher that this would be considered pointless according to the Devs.

    But sometimes you need better compression (better quality at a given bitrate) - it depends what the output goal is

    Most encoders look produce ok results if allocated enough bitrate. But the key is "enough" - that is dependent on content complexity.
    Quote Quote  
  20. So they perhaps use 1pass VBR (Intelligent Recode" and Quality Mode" , 1pass setting), not any form of 1pass Constant quality, and bitrate is based on final resolution and that percentage is just to give it some slack, free play in both directions. If that is true, definitively 2pass would be better choice. But this is just guessing again.
    Quote Quote  
  21. Member
    Join Date: Sep 2004
    Location: australia
    Search Comp PM
    Originally Posted by _Al_ View Post
    So they perhaps use 1pass VBR (Intelligent Recode" and Quality Mode" , 1pass setting), not any form of 1pass Constant quality, and bitrate is based on final resolution and that percentage is just to give it some slack, free play in both directions. If that is true, definitively 2pass would be better choice. But this is just guessing again.
    this is what one of the Devs emailed to me last nght when i enquired about doing a single or double pass for both h264 and dvd mpeg2 outputs, as i am doing some output testing for them and i had to ask him this question.

    As an FYI, if you do a single pass the H.264 encoder was never very good about honoring the average bit rate as it doesn't really know how to allocate bits for future frames. This is a problem will all single pass encoders, some just hide this fact better than others.

    remember, i can set all my outputs to either 1 pass or 2 passes, i use 2 pass for dvd Mpeg2 always, but until now i used 1 pass for 1080/50p to 720/50p

    VRD also has an output mode called "Forced Recode" as well, but was told not to use it.

    he also told me ages ago that for 1080/50p h264 to 720/50p there is no point in using any higher than 45 to 50% of the source files bitrate.
    Last edited by glenpinn; 9th Apr 2014 at 23:15.
    Quote Quote  
  22. Member
    Join Date: Sep 2007
    Location: Canada
    Search Comp PM
    Originally Posted by glenpinn View Post

    As an FYI, if you do a single pass the H.264 encoder was never very good about honoring the average bit rate as it doesn't really know how to allocate bits for future frames. This is a problem will all single pass encoders, some just hide this fact better than others.
    Not entirely true either - some encoders have a "lookahead" function so they do know how to allocate for past / future frames


    VRD also has an output mode called "Forced Recode" as well, but was told not to use it.
    This forces a re-encode of the entire thing - basically makes smart rendering usless

    he also told me ages ago that for 1080/50p h264 to 720/50p there is no point in using any higher than 45 to 50% of the source files bitrate.
    That bold generic statement should be ok for "average" cases, "average" source files - but anything with high content complexity will suffer using their / mainconcept h.264 encoding engine . Even using the higher end mainconcept version with all the bells and whistles shows deficiencies when comparing it to the "gold standard" h.264 encoder (x264). This is easy to proove and demonstrate . The vrd version is even worse than other mainconcept implementations, because it doesn't have control over AQ settings - shadow textures, flat areas end up looking poor
    Quote Quote  
  23. Member
    Join Date: Sep 2004
    Location: australia
    Search Comp PM
    ok so this Mainconcept encoder they are using is not that good, either for dvd Mpeg2 or h264 encoding.

    is it not right that they are adding their own presets that make it better than it would otherwise be, or doesnt it work like that.
    Quote Quote  
  24. Member
    Join Date: Sep 2007
    Location: Canada
    Search Comp PM
    Originally Posted by glenpinn View Post
    ok so this Mainconcept encoder they are using is not that good, either for dvd Mpeg2 or h264 encoding.

    is it not right that they are adding their own presets that make it better than it would otherwise be, or doesnt it work like that.

    It's MPEG2 implementation is good, but it's h264 implementation is average (but x264 is much better, proven time and time again, both in subjective assessment, objective metrics)

    For example, if you took something like the well know test sequence parkjoy at 720p50 24Mb/s (as the "source", to elmiminate resizing algorithm differences, encoded with x264 of course so it looks decent to start with), and re-encoded it x264 and mainconcept at 1/2 the bitrate, there would be a HUGE day/night difference, and that's done with the mainconcept SDK. In fact, I 'm fairly certain I posted some of these comparisions in other threads, I'll try to dig up some old comparisons if you're interested. The reason is there is a lot of motion , dark textures and flat areas (water) , and it taxes compression severely. It separates the men from the boys

    Conversely, if you had say, something like a talking heads shot / interview piece, with very little motion, low content complexity then 11-12Mb/s might even be overkill. Compression is very content dependent
    Last edited by poisondeathray; 9th Apr 2014 at 23:53.
    Quote Quote  
  25. Originally Posted by glenpinn View Post

    As an FYI, if you do a single pass the H.264 encoder was never very good about honoring the average bit rate as it doesn't really know how to allocate bits for future frames. This is a problem will all single pass encoders, some just hide this fact better than others.
    1pass CRF or constant quality, whatever Mainconcept term uses, is just fine, all this time I was just trying to establish if VideoRedo is using that one. Looks like it does not, so sure thing VBR 2 pass is what you should go.

    Not having that mode really sucks , because thinking that you have to encode to certain bitrate is wrong nowadays, just makes no sense, and on the top of that you have to go 2pass. That is why they mask all of this to quality settings, they turn it around for user, at least guessing bitrates for them behind scenes. btw. CRF is such a great tutor for video encoding ...

    but reading pdr, looks like their license version is not top notch anyway ...
    Last edited by _Al_; 10th Apr 2014 at 00:03.
    Quote Quote  
  26. Member
    Join Date: Sep 2004
    Location: australia
    Search Comp PM
    so i am bashing my head against the wall then.

    i still love VRD 1080/50p m2ts output to dvd Mpeg2, no difference to TMPGEnc that i can tell using 2 pass 8000vbr
    Quote Quote  
  27. Then use it , and for that H.264 encoding as well, whatever works for you, just give it enough of bitrate. 720p50 , regular camcorder video, 20 seconds, give it 12Mbit, just do some quick tests, how difficult textures behave, play with it for one hour, and then you will have an idea if it is enough ... Nobody is looking at details, everybody is watching the video, not pixels ...

    Use bitrate viewer, check actual bitrate distribution. What you set and what encoder distributed for scene, or better if there is two quite different scenes (day and low light).
    Quote Quote  
  28. Member
    Join Date: Sep 2004
    Location: australia
    Search Comp PM
    i am almost finished encoding all my files, i will upload them to Mediafire and you can all download them and offer your opinion.
    Quote Quote  
  29. Member
    Join Date: Sep 2007
    Location: Canada
    Search Comp PM
    I'm not going to download them - but can tell you from a lot of experience what to look for.

    Look in shadow areas, there will be detail loss even at "100%" . It's just a problem with mainconcept AVC in general . Their higher end versions help with various AQ settings, but still worse than x264

    The other problems in motion will be keyframe popping and temporal stability issues . The allocation between I,P,B frames isn't ideal

    You uploaded a 1080p50 sample in another thread , I re-encoded it with VRD at 100% quality, intelligent. Look under the roof shadows, that's the typical type of "mushy" shadow texture losses you want to look for. It's not a huge difference here , but on more complex sample, the differences will be striking. The "original" for comparison was resized with spline36 algorithm
    Attached Thumbnails Click image for larger version

Name:	orig.png
Views:	32
Size:	1.25 MB
ID:	24433  

    Click image for larger version

Name:	vrd 100%.png
Views:	13
Size:	978.6 KB
ID:	24434  

    Quote Quote  
  30. Member
    Join Date: Sep 2004
    Location: australia
    Search Comp PM
    well you have better eyes than i do, and i always thought my outputs were fine, many people comment about the video shot with this SDT750 camera.

    i also assumed that h264 and x264 was the same thing, clearly not, so what do you suggest i use, given i edit the video and usually add a title and sometimes an ending credit, which VRD does.

    i downloaded a trial of Sony Movie Studio platinum to see what that is all about, as i also want to add music tracks and probably some little transitions between scene changes to my files if they are needed, and VRD doesnt do that, all i can do is add a text frames between scenes.
    Quote Quote