QuEnc settings for many short low-bitrate DVD files?

[ShawnFumo]

Hi all :) Kind of an interesting situation I've found myself in. I captured like 12 hours of DV footage from an event that I'd like to send to some friends on DVD. I'd used WinDV to capture each cut to its own file to help manage things. It is NTSC and interlaced and most of the shots are indoors (not the best lighting..).

I gave it to a couple of people already as a bunch of XviD files on two DVDs, but a lot of people don't have a DVD-ROM or their computer is too slow, and it is kind of a pain to watch that much stuff on your computer anyway. So, thus begins my efforts to make a real DVD, but using as few DVDs as I can while having relatively decent quality (need to be able to make out the tricks people are doing).

Now, I know that I could try to concat the files into several big files, but several issues with that. First, I can't seem to manage it with Avisynth (just too many files for it to handle at once) and not really enough space on the drive for an intermediate file. Plus, if I have multiple files, I can later drop or re-order files easily (even make multiple paths on the same dvd using dvdauthor, with no additional space required).

But then that makes for an unusual encoding situation. If many of my files are 15 seconds to a minute long each (and usually in the same locale), then should I even bother with VBR and 2-pass encoding? In recent experiments, the clips tended to start blocky and gradually get better near the end, so I'm thinking those settings might even be detrimental in this case?

But then should I just have CBR? And CBR should eliminate the need to control max bitrate right? Or is VBR still something I should do Also, scene detection should only matter for scene changes inside of a video right? So no difference on a small continuous shot?

Right now, 720x480 at 5000kbps looks very nice (but is going to be too large). 2500kbps is breaking up on pans, so that seems out. Half-width (352x480) at 2500kbps is what I'm leaning toward right now. That'd let me get about 4 hours onto each DVD, with 4 DVDs total. I may try going down to vcd resolution and see what that gives me, but it almost feels wrong to have vcd quality on a dvd.. ;)

I'm going to continue experimenting, but any thoughts on the best settings for something like this? Also, from a recent thread, it sounds like maybe I should grab an older version of QuEnc since I'm using a lower bitrate? Does it really make that much difference?

Thanks,
Shawn

[ShawnFumo]

Well.. having CBR seems to be working pretty well. I compared with the old VBR files and found that the average bitrate on a lot of them was actually less than I'd specified. I don't think QuEnc has a minimum bitrate option, though. Overall, the CBR files seem to look better.

I experimented with VCD size and real VCD (mpeg1), but I couldn't get any of it to look very well. Does VCD always produce combing on pans? It seems that even if I just resize down and have it look blended on the computer, it still combs as it pans on the TV. After thinking about it a while, I guess it must be from the TV applying its own interlacing to progressive half-height source?? Maybe I'm just doing something wrong, but I tried all kinds of things... I think I'll stay with half-width DVD for now. ;)

Installed 0.54 of QuEnc, but then I remembered that the old versions couldn't encode and mux the audio. I guess I'll still try making some video streams and see how much better they look. I'm using batch files anyway, so I it isn't like I can't add in the various other steps, but it really seems so much more elegant just to do it all in QuEnc. :)

Still, if anyone thinks that VBR and/or multi-pass can be made to work better in my situation with the right settings, let me know.

Cheers,
Shawn

[drcl]

CBR is never better than VBR, cant think of how you are getting these results.

Max bitrate isnt an issue for DVD when your average is so low anyway.

The biggest problem you are probably having is noise, which is lowering the quality of the final encoding.

vcd doesnt produce combing and the tv doesnt introduce interlacing. vcd is half height, so the source you feed in to the encoder should be progressive. Cant understand how you are getting combing if you did a 2:1 resize. When you play back the clip on the computer make sure the mpeg deinterlacer is on weave.

can you post your Quenc settings?
and your avisynth script?

I havent heard of AVISynth having a file limit. What error does it produce?

Short clips may screw up the bitrate control. It would be better to group the clips in to several larger ones by theme or something. You will get beter bitrate distribution and better quality. The best solution would be to encode each disk as one long clip then break it up in to smaller parts to optimise the bitrate. A lot of work and I wouldnt know how to do it well.

Quenc seems to miss the bitrate by about -50kbits/sec; simply add 50 to your specified bitrate.

Also with quenc 0.54 you need to be wary of the field order as there is no choice. Easily corrected using Restream after encoding.

[dragongodz]

i would go with the half D1(what you call half width) myself. should be pretty good quality.

CBR is never better than VBR, cant think of how you are getting these results.
ofcourse QuEnc doesnt actually do CBR because of the still iffy rate control. the CBR setting should be much more constrained though and less variable than the VBR setting.

[Abond]

Originally posted by drcl


Max bitrate isnt an issue for DVD when your average is so low anyway.

Hmm, not completely true. Avg 2500 with 8000 max give me over 12000 peak with some files - not acceptable for me.

[drcl]

You could try HC(independent of avcodec) or FreeEnc.

FreeEnc allows you to adjust the rate control algorithm I believe.

12000 on a 2 pass? I've never seen that. Even on high action at 2500 it rarely goes above 6000 with 0.54

What material is that?

@shawnfumo

12 hours indoors? music festival? If it is something like that you could use a good deinterlacer (KernelDeint) to give a film like effect and allow progressive encoding which will lower the bitrate requirement a fair bit.

[Abond]

Hmm, well I really don't need advice :) but thank you anyway. I am using HC quite often and some times FreeEnc which has the same flaw - over max bitrate (quite OVER).
As for the material in all cases tested with DVD source (full encode) and MPEG2 TV capture (movies, also full encode).
Hmm, that is a bit misunderstanding, I said nothing about 0.54 (this is completely another case with undersized files, and correspondingly lower than setted average bitrate IIRC). Target in all cases - 720X576 DVD.

[drcl]

my post was a little confusing. various ppl had bought up possible issues and problems in relation to the original post, which i was trying to address. I should have done it properly with lots of "@"s but I'm new to posting and i was in a rush. sorry

0.54 was mentioned by shawnfumo.

I wasnt giving advice to you. The "you" i was refering to was actually shaunfumo but also the general audience, ppl who may have the same problems, for that is how I think of forums.

As to material, I meant what was the actual type of footage creating the high bitrate?

[ShawnFumo]

Originally posted by drcl
CBR is never better than VBR, cant think of how you are getting these results.


I'll have to post some of my encodes and such. They're small files anyway for the most part.


The biggest problem you are probably having is noise, which is lowering the quality of the final encoding.


Well, noise probably isn't helping. I've been trying to avoid having to learn all this crazy convolution and temporal smoothing and all of that. ;) Plus, I'm used to encoding to XviD, which seems to be good at dealing wit noise itself usually...


vcd doesnt produce combing and the tv doesnt introduce interlacing. vcd is half height, so the source you feed in to the encoder should be progressive. Cant understand how you are getting combing if you did a 2:1 resize. When you play back the clip on the computer make sure the mpeg deinterlacer is on weave.


Yeah, I'm not sure what's up with that. The funny thing is that it looks OK on the computer (even looking frame by frame in virtualdubmod), but then weird on the TV when panning. Maybe it is because I'm using my DVD player to play it as an mpeg on a cd-r instead of as vcd disp per se, but was doing the same thing with 352x240 mpeg2 as well..

But I think I may just stick with the half D1 anyway (thanks dragongodz for the name). I think that's the only way I'll have the full smoothness (semi-60fps thing) of DV anyway and no blended images with pausing on a TV. If it is the difference between 3 and 4 DVDs, may as well go all out.. :)


I havent heard of AVISynth having a file limit. What error does it produce?


With AVISource, it gives an error of not having an available decompressor for dv (well, whatever exact 4cc is). I'm thinking it is probably some limit on VfW, in that Avisynth is probably asking to have all these files open at once, even if it is just one frame at a time? I think it got up to over 100 files. Directshowsource got farther (though took a looong time) but also finally gave an error. I'd have to check on the exact one. Maybe some sort of plugin is needed that will take a big mass of files listed in a text file and treat it as one entity which only opens on VfW instance at a time...


Short clips may screw up the bitrate control. It would be better to group the clips in to several larger ones by theme or something.
...
A lot of work and I wouldnt know how to do it well.
...
Quenc seems to miss the bitrate by about -50kbits/sec; simply add 50 to your specified bitrate.


Yeah... I think the shortness is really confusing things a bit. I'll try just bumping up the bitrate, but I'm not sure if it is off consistantly or not. If it IS off consistantly, that could work out pretty well.
[/quote]


Also with quenc 0.54 you need to be wary of the field order as there is no choice. Easily corrected using Restream after encoding.

OK, I'll keep that in mind if I do try 0.54.

[ShawnFumo]

drcl: Thanks for mentioning HC. I've never actually seen that one mentioned before, but it sounds like it has a lot of options. I'd originally passed over FreeEnc because of QuEnc's muxing capabilities, but I'll have to revisit that.

Music festival was a good try, but it was the World Yo-yo Contest. :) The 12 hours of footage was spread out over the night I got there and the three main days of the contest, workshops, BBQ, etc. Most of that was in a hotel/convention thing in Orland, but now that I stop and think of it, there was a fair amount of outdoor footage on the day of the BBQ. And even in the hotel, there was variation between the main ballroom versus the hallways versus the freestyles on stage, etc. Not to mention the airport and on the airplane.

Geeze, I guess it won't be all that consistant after all. I've been trying to put the content out of my mind, since I still need to make a clip video at some point (probably multiple more than one really). Maybe I can do some retrospectives before this year's contest in July. Anyway, now I'm just rambling.. ;)

Thanks for the advice guys,
Shawn

[Boulder]

Originally posted by ShawnFumo
Yeah, I'm not sure what's up with that. The funny thing is that it looks OK on the computer (even looking frame by frame in virtualdubmod), but then weird on the TV when panning.

You might have set an incorrect field order. DV is bottom field first if I remember correctly. Encode as interlaced with BFF order set. If you can't set the field order in the encoder, try DoubleWeave().SelectOdd() in the script, it will reverse the field order to TFF.

You really should denoise the video, it will help a lot. There are lots of threads in the Avisynth usage forum regarding denoising interlaced content. Pay attention to scharfis_brain's posts, he knows the stuff very well.