Batch encode with 2-pass Xvid

[Wam7]

I normally use VDub Mod and Xvid 2 pass for my conversions to watch on PocketPC but I have been asked to do some conversions that may involve 100's of files and setting up 2-pass encode, demuxing the audio and converting separately etc will mean much more time.

I tried Super which does a good job but does not do 2-pass encoding which gives me much better quality (which allows smaller file size). Is there another app to to convert a large batch of files with 2-pass Xvid in relatively easy steps?

[Soopafresh]

Download Winmenc http://yawoogle.googlepages.com/winmenc.0.61.zip

Download MPlayer http://oss.netfarm.it/mplayer/builds/mplayer-p4-svn-22220.7z

Download Mencoder http://oss.netfarm.it/mplayer/builds/mencoder-p4-svn-22220.7z

Download PthreadGC2 ftp://sourceware.org/pub/pthreads-win32/dll-latest/lib/pthreadGC2.dll

Place all in the same folder. Run Winmenc.


This will create Xvid Divx Mpeg1 Mpeg2 and h264 (in avi container), all 2 pass options available.

[Wam7]

Thank you very much Soopafresh! Just doing a conversion now and it seems just the thing I needed. Should also prove handy info to others.

[jagabo]

You can save time and get just as good results by using 1-pass target quantizer mode. You won't be able to predict the exact file size beforehand though.

With 2-pass VBR encoding you know the file size but you don't know the quality. With 1-pass CQ encoding you know the quality but you don't know the file size.

[Wam7]

You can save time and get just as good results by using 1-pass target quantizer mode. You won't be able to predict the exact file size beforehand though.

With 2-pass VBR encoding you know the file size but you don't know the quality. With 1-pass CQ encoding you know the quality but you don't know the file size.

Encoding for handheld devices I need to keep the file size pretty constant (the smaller the better but have to balance quality with size) and a 2-pass encode is normally always better than a 1-pass when the file sizes are the same.

I have tested this method out and not only is the quality just as good but also the encoding times are much quicker. In tandem with the fact that I don't have to 'do it by hand' this has saved me so much more time. Thanks again Soopafresh.

[Soopafresh]

Glad to help out

[jagabo]

You can save time and get just as good results by using 1-pass target quantizer mode. You won't be able to predict the exact file size beforehand though.

With 2-pass VBR encoding you know the file size but you don't know the quality. With 1-pass CQ encoding you know the quality but you don't know the file size.

...a 2-pass encode is normally always better than a 1-pass when the file sizes are the same.

I disagree. I have very often compared single pass constant quality (not constant bitrate) encodes with 2-pass encodes of the same size and the quality is almost exactly the same. I've done this with Xvid, Divx, TMPGEnc, CCE and X264. If you look at enlarged still frames you can see very minor differences here and there but sometimes one will be better, sometimes the other.

[PuzZLeR]

Multipass would not exist with its longer encode times if it wasn't a good option for quality as well.

Here's the best solution IMHO. It's an extra step in encoding, but worth it.

Encode first with quality based. After, use an app like VDub to determine average bitrate used that gave this certain level of quality and WRITE IT DOWN.

Encode again, this time with Multipass (from same source obviously), but use the average bitrate you determined with quality based in previous step.

Compare.

I find the quality is always a bit better with Multipass. If you know in advance what bitrate you need to ensure good quality, multipass will optimize it by spreading it to the places that need it most for a given file size.

I would always delete the quality based files after I was done. Now I actually delete them after I write down the average bitrates even before I do another encoding because I know I won't need them again.

In fact, I have now moved on to H.264/AVC with Nero Recode but I still run a quality based encoding run with DivX to get an idea of the bitrate I would need for Recode's Quality based 2-pass.

Yeah, Recode has quantizers too, and gives similar data, but currently H.264/AVC is still slow to encode so I actually use DivX instead, for this first step, just to gather my data. Actually I'm currently expirementing with Recode's MPEG-4 ASP quantizers since it encodes faster for my data, but not sure of the accuracy yet.

[jagabo]

Multipass would not exist with its longer encode times if it wasn't a good option for quality as well.

Nobody said it wasn't a good option for quality. Multipass is necessary when you want maximum quality for a given file size. Like when putting 4.3 GB of MPEG2 video on a DVD. Or 700MB of Divx video on a CD.

Here's the best solution IMHO. It's an extra step in encoding, but worth it.

Encode first with quality based. After, use an app like VDub to determine average bitrate used that gave this certain level of quality and WRITE IT DOWN.

Encode again, this time with Multipass (from same source obviously), but use the average bitrate you determined with quality based in previous step.

Compare.

I find the quality is always a bit better with Multipass.

That's exactly what I've done in my testing. And I always find the quality is the same. As noted earlier, you'll find some frames in one that look better, some in the other. In the end they average out. Guess we'll just have to disagree.

[PuzZLeR]

Multipass is necessary when you want maximum quality for a given file size. Like when putting 4.3 GB of MPEG2 video on a DVD. Or 700MB of Divx video on a CD.

Sure. How about the file size determined from quality based encoding's final average bitrate? Wouldn't that too be maximum quality for a given file size (since average bitrate is directly related to file size)? Maybe flawed, but that is how I logically see it.

Guess we'll just have to disagree.

Hey no problem, I respect that jagabo. I am only basing my opinion on only my eyesight and my tastes no doubt. This could easily be a debate that can make this thread go many pages.

In fact, I too do notice some Q-based frames do come out sometimes nicer, but not enough percentage of them in my opinion to drop the much longer multipass encoding step.

I think a better utility would be one that inputs several video clips, and with some sort of technical testing, chooses the best frames from each, or the frames closest to the source in quality (given the source as a benchmark) and outputs a final optimized video clip.

That would be neat.

[jagabo]

Multipass is necessary when you want maximum quality for a given file size. Like when putting 4.3 GB of MPEG2 video on a DVD. Or 700MB of Divx video on a CD.

Sure. How about the file size determined from quality based encoding's final average bitrate? Wouldn't that too be maximum quality for a given file size (since average bitrate is directly related to file size)? Maybe flawed, but that is how I logically see it.

Of course you could use single pass quality based encoding to get the best quality for a specific file size -- but you'd have to perform the encoding over and over again to find the quantizer that gives the size you want. You might as well do a multipass encode and save the time and trouble.

[PuzZLeR]

What Q-Based does for me is provide me with an idea, and a comparison, between many different files. If some files end up with a higher bitrate then this is telling me that this clip may have more motion and therefore has a demand for higher bitrate.

If you want a specific file size, yes, multipass gives it to you. But in reality, whether Q-based or multipass, virtually every video clip will have different bitrate needs for a certain level of quality. A "one-bitrate formula" for multipass may simplify things, and helps in budgeting video length and/or disc space, but you will never optimize quality this way. You will either end up with lower quality or too much bitrate for what you want this way.