Xvid Resizing

[Cunhambebe]

You may be fed up with this subject but I'd like to know something important - resizing
I'm used to converting to Xvid (with VirtualDubMod) my videos previously rendered with Lightwave (MPEG2 at 720x480). As I convert them, I'm used to resizing the video to 640x480 using full processing mode (drawback - loss of quality, but it's the only way to make VD filters work - the filter "resize"). I want to let people get my scenes on the web and re-convert them again as MPEG2, resizing the video back to 720x480. Am I right resizing the video in VirtualDub as I convert it to Xvid? Would a filter null transform be better or I should leave video resolution as it is as I convert to Xvid?
thanks in advance

- PS: you can see some of my animations at
http://www.lwg3d.org/forums/forumdisplay.php?f=229
- you don't need to register to see them.

[Soopafresh]

I think you will get better resize quality with Avisynth using the line

Lanczos4Resize(720,480)

or whatever dimension you choose.


In conjunction with a commandline app like avs2avi, you'll have a quickly resized avi file.

[bugster]

I know nothing about Lightwave, but can it not output directly to .avi of some kind? Maybe even the xvid/divx you want for the web? This would remove one re-encoding step and help to maintain quality.

As for resizing and maintaining quality, I would reccomend you encode your avi's as 720 x 480 if you expect them to be re-encoded to DVD with the best possible quality. Re-encoding is generally not a good ide as you knwo, adding in resizing is just going to increase the quality reduction.

An avi at this resolution may look as if the aspect ratio is wrong when displayed on your PC, but this would be corrected when encoded to DVD.

If you have the space and time, maybe you could offer both resolutions. 640 x 480 for the PC viewer, 720 x 480 for those that want to convert to DVD.

[junkmalle]

Quality loss from the YUV/RGB colorspace conversion is minor compared to the all the other losses you get during XVID compression at the bitrates you're using.

VirtualDub's Null Transform filter is exactly that. It doesn't do anything to the video. It's only there to allow cropping in the absence of other filters.

[celtic_druid]

Lightwave can indeed output to avi. Personally I would output an image sequence though (which I already suggested) that way if there is a crash or anything you still have all the frames already rendered, plus you can use multiple machines for rendering.

[Cunhambebe]

Thanks to all for the valuable help.
Soopafresh:
-Thanks for the input. I'm sure AviSynth is great - but I'm stil trying to fix this before trying AviSynth - and I'm sure it's gonna be hard to learn it. Anyway, you stated that it's better to keep 720x480, so I understand your advice is to keep 720x480?

bugster:
I know nothing about Lightwave, but can it not output directly to .avi of some kind? Maybe even the xvid/divx you want for the web? This would remove one re-encoding step and help to maintain quality.
- people who have Lightwave, neither render their scenes as AVI nor as .MOV since if something happens while you're rendering, you'll loose everything and have to start it all over again from the very beginning. So, if you are planning to render something in Lightwave you'd better choose an image sequence as celtic-druid well stated - for instance, as Targa 32 (this way if something happens....all your previous frames are already saved), NTSC (if you intend to show your work on regular tv) at 720x480 or 720x486. After the sequence is ready, you can drop it in any MPEG encoder and render the final thing as MPEG2 DV 720x480 (sorry not to make this clear right overthere). Point.

(...)Re-encoding is generally not a good ide as you knwo, adding in resizing is just going to increase the quality reduction. (...) An avi at this resolution may look as if the aspect ratio is wrong when displayed on your PC, but this would be corrected when encoded to DVD.
- I know that, but anyway, keeping the aspect ratio at 720x480 will make the Xvid file much heavier for slower computers and it will certainly look odd on PC monitors,a s you stated right above (black stripes on the top and at the bottom of the screen). Here's an interesting question: I wonder what kind of video size people get when they download them from the web - and this is something no one can deny.

If you have the space and time, maybe you could offer both resolutions. 640 x 480 for the PC viewer, 720 x 480 for those that want to convert to DVD.
- that's a good idea and a bad one at the same time- lol. Although space is not an issue where I upload my scenes, here at home it certainly is.

junkmalle
Null Transform filter is exactly that. It doesn't do anything to the video. It's only there to allow cropping in the absence of other filters.
- Here you already have a quality loss when you choose Full Processing Mode to make available VD filters such as null transform. Anyway, I've already tried turning on this filter (only this one) as I convert it to Xvid. Nothing seems to happen. Final file look as though it was coverted at full size, I mean, dark stripes on the top and at the bottom of the screen. (?) So, what? - lol

celtic_druid:
- Yes, you're right, but how about resizing???? What would you suggest?
Thanks to all, but to make this discussion concise, please I'd like to know what I should do!!!

[guns1inger]

Don't cater for the resize crowd. Put them up on the website in a convenient size for downloading and viewing online, then sell the DVD versions to those who want full size. There is not way you can resize, heavily compress, then expect someone at the other end to return it to it's original state.

[Cunhambebe]

I don't sell anything (I guess I should - it's hard to be a lawyer in Brazil - lol lol - just kidding). I do it for fun and for educational purposes (I'm serious here). I believe in perfection and that means, since you decide doing a certain task, you must do it right to achieve...perfection Anyway, thanks for your tips.

[bugster]

Cunhambebe
, thanks for the explanation of Lightwave. So now I say, if you want maximum quality for your avi's, encode straight to avi from the original lightwave output (Targa sequence or whatever). I can't reccomend an app that will do this, but I will be very surpised if you can't find one that will accept your image sequences as input for avi encoding.

As to resizeing, I still say for max quality, leave the avi at 720 x 480. It won't be a problem for PC playback on slower computers compared to 640 x 480. Also, if someone wants to view it on a PC at the correct aspect ratio then they simply need to use a player that allows them to resize on the fly. As long as you make this info available to the downloaders they have nothing to complain about!

[Cunhambebe]

Thanks, bugster.
I can't reccomend an app that will do this, but I will be very surpised if you can't find one that will accept your image sequences as input for avi encoding.
-Lightwave itself can do that. You may drop the sequence in Lightwave as a background, hit image editor and select sequence and render the scene once more - but this time as AVI.

Also, if someone wants to view it on a PC at the correct aspect ratio then they simply need to use a player that allows them to resize on the fly.
- I'd like to know if you can suggest one....How about Media Player Classic. Does it do that?

Thanks for your help

[celtic_druid]

If you are encoding with XviD. Just set the PAR when encoding and it should be resized on playback.

VirtualDub/Mod accepts images as input, as does AVISynth.

[junkmalle]

junkmalle
Null Transform filter is exactly that. It doesn't do anything to the video. It's only there to allow cropping in the absence of other filters.
- Here you already have a quality loss when you choose Full Processing Mode to make available VD filters such as null transform. Anyway, I've already tried turning on this filter (only this one) as I convert it to Xvid. Nothing seems to happen. Final file look as though it was coverted at full size, I mean, dark stripes on the top and at the bottom of the screen. (?) So, what? - lol

Sorry, I thought the part about losses from the YUV -> RGB transform was obvious. What I meant was, beyond that, Null Transfrom does absolutely nothing to the video. There is no reason to use it except to crop the video (because there is no explicit crop filter; cropping is applied to any other filter). You can add Null Transform a million times in a row and there will be no difference between the input to the first one, and the output of the last one.

I'm guessing you are starting with a series of RGB images (TGA, BMP?). In this case, enabling full processing mode has no effect on video quality at all: there is no YUV -> RGB conversion, your images are RGB already.

Most your losses are coming from subsequent handling. Some will come from the RGB -> YUV colorspace conversion within XVID, more will come from the 4:1:1 sample reduction, and at the bitrates you're using most will come from the compression itself. So there's no point in worrying about Full Processing vs. Fast Recompress.

Even if you are starting with a YUV source (JPEG series?) the initial loss going from YUV to RGB is minimal compared to losses from Xvid compression.

[Cunhambebe]

Thanks, as always...
celtic_druid
If you are encoding with XviD. Just set the PAR when encoding and it should be resized on playback.
- Thanks for the input. I'm sorry, but you "may be" wrong here, setting PAR as default doesn't resize anything automatically on playback- I've just tried 4:3 - result is the same: no resizing at all. Bugster's remark on resizing on the fly was great too! I've just checked Media Player Classic. It can resize it on the fly - what a great player.
VirtualDub/Mod accepts images as input, as does AVISynth.
- You mean, both can deal with sequence files? If the answer is yes, I guess I can drop my sequence files (targa 32) in VD, convert it as Xivd, use Vegas to make the soundtrack, render this as AC-3 and then mux Xvid + AC-3 with Avi Mux Gui. Hmm, interesting. Hope this is possible.
junkmalle
- thanks so much for the detailed explanation.
(...)What I meant was, beyond that, Null Transfrom does absolutely nothing to the video. There is no reason to use it except to crop the video (because there is no explicit crop filter; cropping is applied to any other filter). You can add Null Transform a million times in a row and there will be no difference between the input to the first one, and the output of the last one.(..)
- Because null transform applies cropping to other filters, right?
I'm guessing you are starting with a series of RGB images (TGA, BMP?). In this case, enabling full processing mode has no effect on video quality at all: there is no YUV -> RGB conversion, your images are RGB already.
-Not at all. I'm converting a final format - MPEG2 to Xvid. Now I know about this possibility (converting sequence files as Xvid - hope I'm right here) as it's been well explained by celtic-druid right above.
Most your losses are coming from subsequent handling. Some will come from the RGB -> YUV colorspace conversion within XVID, more will come from the 4:1:1 sample reduction, and at the bitrates you're using most will come from the compression itself. So there's no point in worrying about Full Processing vs. Fast Recompress.
- I don't know. Actually, I'm converting MPEG2 to Xvid. If you want to apply some of VD filters you must choose Full Processing Mode (yes, there's a quality loss here). No filters, no resizing, then you can choose Fast Recompress (no quality loss). Also there's another possibility: you can still use VD and choose Direct Stream Copy to frameserve the output to another application such as Vegas and there, convert the MPEG2 as Xvid - of course this makes sense only if you're using some of VD's filters.

Interesting note here: the inverse way
When you reconvert the final Xvid as MPEG2 you can frameserve from VD to any other MPEG encoder such as TMPEGEnc, MainConcept Stand-Alone or even Vegas(it has a built-in MainConcept encoder, even though letting Vegas deal with Xvid files, first you must change Xvid's FourCC with a tool such as AVI FourCC Code Changer - and I don't know what to choose among all those options there).
Even if you are starting with a YUV source (JPEG series?) the initial loss going from YUV to RGB is minimal compared to losses from Xvid compression.
-Ok, let's say I'm not resizing anything in VD (I'll let the others resize th efile on the fly) and that I'm working with MPEG2s. This way I must choose deinterlace (since my MPEG2s are interlaced - intended for regular TV as any other VOB). Therefore, Full Processing Mode is an obvious choice. Is the loss noticeable, despite what you've stated (...the initial loss going from YUV to RGB is minimal compared to losses from Xvid)? I've converted some files here and Full Processing Mode outputs files just a little blured. Another option would be converting sequences and dealing with the soundtrack in Vegas as explained above.
Many thanks to all.

[GreatScott]

http://www.vrac.iastate.edu/~chadspen/interlace/xvid-1.0.2-mod-dec.exe.bz2

[junkmalle]

[/u]I'm sorry, but you "may be" wrong here, setting PAR as default doesn't resize anything automatically on playback[/u]- I've just tried 4:3 - result is the same: no resizing at all.

PAR is a kludge bolted on to the AVI spec. Most players don't understand PAR.

You mean, both (VirtualDub and AVISynth) can deal with sequence files?

Yes.

Because null transform applies cropping to other filters, right?

No. VirtualDub can only apply cropping to a filter, any filter. Null Transform gives you a filter that does nothing to your video so you can crop in the absence of other filters.

Actually, I'm converting MPEG2 to Xvid. If you want to apply some of VD filters you must choose Full Processing Mode (yes, there's a quality loss here). No filters, no resizing, then you can choose Fast Recompress (no quality loss).

You're still going to get quality loss from the XVID compression.

Ok, let's say I'm not resizing anything in VD (I'll let the others resize th efile on the fly) and that I'm working with MPEG2s. This way I must choose deinterlace (since my MPEG2s are interlaced - intended for regular TV as any other VOB). Therefore, Full Processing Mode is an obvious choice. Is the loss noticeable, despite what you've stated (...the initial loss going from YUV to RGB is minimal compared to losses from Xvid)?

The damage done by deinterlacing is 100 times worse than the YUV -> RGB -> YUV conversion.

Try this: convert one of your MPEG files to uncompressed RGB AVI (full processing mode, no deinterlacing). compare the results to the original. I bet you won't be able to tell the difference.

Another option would be converting sequences...

That's probably your best bet, especially if your sequence files are not interlaced.

[Cunhambebe]

Hi there. Thanks so much for your help, junkmalle
PAR is a kludge bolted on to the AVI spec. Most players don't understand PAR.
- lol - Klugde what? Oh boy.... I know what you mean but I though celtic_druid was talking about one of those options for Pixel Aspect Ratio when compressing with Xvid.

Null Transform gives you a filter that does nothing to your video so you can crop in the absence of other filters. (?) - lol

Try this: convert one of your MPEG files to uncompressed RGB AVI (full processing mode, no deinterlacing). compare the results to the original. I bet you won't be able to tell the difference.
- you're telling me not to use deinterlace?

Another option would be converting sequences...
That's probably your best bet, especially if your sequence files are not interlaced.
- I've just tried converting a sequence. Quality is just unbelievable- fantastic -Targa is progressive!!!! I know I can frameserve from Vegas to VirtualDub (Debugmode FrameServer Plug-in), or convert the whole sequence to Xvid n Vegas, but I'd like to know if there's any way to join (edit) "sequences" in VirtualDub. I've read some tutorials on the web and most of them refer to AVI files, so the option <Apend AVI segment> seems not to be availabe for sequence files. Any ideas? Thanks again!!

Just in time: hey, thanks GreatScott! Is that a filter or what?

[junkmalle]

I'd like to know if there's any way to join (edit) "sequences" in VirtualDub.

[edit]
Oh, sorry, I think I misinterpreted what you were asking. You want to join different sequences of TGA files in VirtualDub. You can't do that unless you want to rename all your TGA files. But if you create AVI files with the same specs you can use VirtualDub to join the AVI files together. That is what I described the first time.
[/edit]

If your AVI files are all the same frame size, frame rate, codec, same audio type and parameters you can join in VirtualDub using the File -> Append AVI Segment function. Use File -> Open to get the first one, then File -> Append AVI segment to join subsequent files. If you're AVI files are numbered (eg, file01.avi, file02.avi...) VirutualDub can automatically load the sequence. Do all this in Direct Stream Copy mode and you will get no quality losses.

[Cunhambebe]

Hi junkmalle, thanks again for your help...
Oh, sorry, I think I misinterpreted what you were asking.
- Nah, no problem. In fact, I'm the one who's got to thank you for everything. So here is it: thanks so much, my friend.

You can't do that unless you want to rename all your TGA files.
-What a pity!

But if you create AVI files with the same specs you can use VirtualDub to join the AVI files together.
-Good idea, even though this way I'm going to create big AVI files (that of course I can delete later). You know, as I try to convert the sequence with Vegas (Xvid as the codec), I notice the final file is much bigger than the one converted with VirtualDub (same settings for compression options). I was wondering why?
Thanks again.

[junkmalle]

You should be able to frame server from several instances of VirtualDub to Vegas. That way you won't have to create a bunch of intermediate AVI files.

Regarding files sizes after conversion: if you are really using the exact same settings for the audio and video the final file should be the same size -- assuming the same source.

Are you comparing AVI files created from MPEG files to files created from TGA files? If you are using quantization (constant quality) mode in XVID that could cause a difference. MPEG may have done something to the video that makes it harder or easier to compress.

I have seen XVID get "stuck" where changing the bitrate doesn't make a difference in the output. The only solution I've found so far has been to uninstall the codec and reinstall it.

By the way, if you don't need files of a particular size I recommend using Quantization mode in XVID. Pick a Quantization value that gives you the quality you want and encode in a single pass. Every frame will have that quality.

I usually use a value between 2 and 4. At 2 you can barely see differnece between the source and destination even with still frames. At 3 you can see minor differences in still frames but not at normal playback speed. At four you can see minor differences at normal playback. You can use decimals. So 2.5 or 3.25 are valid values too.

[celtic_druid]

Wrong. The PAR info is indipendant of the avi container. It is part of the MPEG-4 bitstream.

XviD, ffdshow, mplayer and VLC all support auto resizing. For XviD you need to be using a 1.1.x build.

Being indipendant if you remux your avi to say mp4 then the PAR info is kept. Same is true of other containers, although with mkv it will be duplicated.

[Cunhambebe]

Thanks to all, as always...
First of all, I'd really like to know what's that file right overthere...GreatScott, would you please tell me where and how I isntall that file? What's it for? Deintelace? Thanks...

junkmalle:
You should be able to frame server from several instances of VirtualDub to Vegas. That way you won't have to create a bunch of intermediate AVI files.
-I've just installed that reg file "proxyon". Let's see if it works this time.

Regarding files sizes after conversion: if you are really using the exact same settings for the audio and video the final file should be the same size -- assuming the same source.
- That's definitely weird, insn't it? I converted a file with VD (created 2 AVIs out of 2 sequences>AVI apend>etc...converted to Xvid with VD). I dropped the same sequences with Vegas (original Targa - no AVIs here), converted to Xvid with the same settings and the result is alwyas a bigger file. What a mystery!

Are you comparing AVI files created from MPEG files to files created from TGA files?
- Yes, Sir.

If you are using quantization (constant quality) mode in XVID that could cause a difference. MPEG may have done something to the video that makes it harder or easier to compress.
- Well, as far as I know, I am using trellis quantization and quantization restrictions 2, 31, 2, 31, 2, 31. Is that what you mean?

By the way, if you don't need files of a particular size I recommend using Quantization mode in XVID. Pick a Quantization value that gives you the quality you want and encode in a single pass. Every frame will have that quality. I usually use a value between 2 and 4. At 2 you can barely see differnece between the source and destination even with still frames. At 3 you can see minor differences in still frames but not at normal playback speed. At four you can see minor differences at normal playback. You can use decimals. So 2.5 or 3.25 are valid values too.
- Thanks so much for the input (hope I'm doing this right 2, 31, 2, 31, etc.

celtic_druid
Wrong. The PAR info is indipendant of the avi container. It is part of the MPEG-4 bitstream.
- Then please, how do you configure this option....I'm a newbie and it's hard to find all these options.....Thanks for the input.

[junkmalle]

Quantization is a single pass option. Just go to the XVID Configuration dialog, set Encode Type to Single Pass, Then, if the button below Encode Type says Target Bitrate press it to make it say Target Quantizer. Then set the value next to the button to set the desired Quantizer. The lower the value the higher the quality.

I tried using XVID's Pixel Aspect Ratio and Picture Aspect Ratio's settings. You can find them by pressing the More button to the right of Profile @ Level setting in the XVID Configuration dialog. Then select the Aspect Ratio tab.

I was running XVID version 1.0.3 -- it had the aspect ratio settings but they made no difference on playback. I updated to 1.1 beta and found that some players do respect the aspect ratio setting now. MPLAYER2.EXE (Microsoft's old Media Player 6.4) used the AR setting. Media Player Classic did not (I may not have the latest version). Divx's player did not.

[Cunhambebe]

This is wonderful. I've just installed the aviproxy (or handler) and neither Vegas nor MainConcept Stand-Alone cannot open the **.vdr file (even renaming the extension as AVI). Vegas pops-up error messages and MainConcept renders nothing but a green screen. Here I'm trying to convert from Targa sequence or AVI to Xvid. Please, what am I doing wrong????
On the other hand, trying to convert Xvid with MainConcep to MPEG2 (inverse way)....it works by simply droppig the Xvid file in MainConcept. As discussed around here before, Vegas seem not being able to work with Xvid files (guess you have to change Xvid's FourCC).

[Cunhambebe]

So being a Master makes all the difference, isn't it? Thanks for the input, junkmalle..
Quantization is a single pass option. (...) The lower the value the higher the quality.
-I see..Do you really think it's better than the 2 pass encoding option?

I tried using XVID's Pixel Aspect Ratio and Picture Aspect Ratio's settings. You can find them by pressing the More button to the right of Profile @ Level setting in the XVID Configuration dialog. Then select the Aspect Ratio tab.
- I knew it.....I was just a bit confused by celtic_druid's statements... I guess I'll leave PAR as default....now, how about the frameserving problem???

[Cunhambebe]

By the way, if any of you are curious to see the Xvid file encoded as it's been discussed here, please go to
http://www.lwg3d.org/forums/showthread.php?t=24404
...hope you like it.
Now please, let's fix the problem with frameserving in VirtualDub....Any tips??

[junkmalle]

Did you run the AUXSETUP program that came with VirtualDub? You need to do that before it can frame serve.

[Cunhambebe]

Thanks as always...
Yes, I did. And I've also installed that reg file that comes along in one of those VD folders. I've noticed that after installing all those files, as I try to convert/encode any file as Xvid (with VD or even Vegas), if I try to delete it, my OS doesn't let me do it. I sounds as VD, even turned off, is frameserving without being requested. Uninstalling those files have fixed the problem. Does any of you have a simple guide on how to frameserve from VD? I've already read all of those around here. PS: thre's just one thing I haven't tried yet: installing a certain dll file....Is that safe?
Thanks in advance.

[Stormin Norman]

Here is some tips for xvid.

Use two pass encoding without 1st pass dicard

Encode 1 pass at a time. Then make your calculations (dont go below 70% of first pass in file size on second pass)

Use perfect xvid(google it) tool for tuning your agression curves max and min% on second pass.

Use lower I and P frame quants on second pass, try 2,6,2,6.

Tons of guides out there, try and stick with the 2 pass guides.

[junkmalle]

I've never had any problems frameserving from VirtualDubMPEG2 or VirtualDubMod. I don't have Vegas so I don't know if there's something tricky with it. Try these links:

http://forum.doom9.org/showthread.php?s=&threadid=47483

http://neuron2.net/LVG/frameserving.html

[Cunhambebe]

Thank you very much, Norman and junkmalle. I'll chek it out. Let's see if I can fix this.
Stormin Norman
Use two pass encoding without 1st pass dicard
- I do.

Encode 1 pass at a time. Then make your calculations (dont go below 70% of first pass in file size on second pass)
- I know I must encode 1 pass at a time. Now, about calculations, how do you calculate this, pressing that button next to 2nd pass?

Use perfect xvid(google it) tool for tuning your agression curves max and min% on second pass.
- I've never heard of.... In fact I've found 2 sites on the web, one in Spanish and the other one in French. Despite the fact I don't speak Spanish (I can read just a little), this site seem to be very complete with guides etc...

http://www.divx-deux.com/espanol/xvid.html

If you don't understand Spanish you can try this one en français:

http://divisk.free.fr/downloads/downloads.php

On the other hand, agression curves...I guess this is somehting too much beyond my knowlegde (but I'll try ! - I never give up)

Use lower I and P frame quants on second pass, try 2,6,2,6.
Tons of guides out there, try and stick with the 2 pass guides.
- How about 2, 31, 2, 31, 2, 31?
Thanks again!