How to convert 4:3 NTSC to 16:9 also NTSC ? and vice-versa ?

[Zetti]

Hi all,

I have two MPEG2 files, extracted from two DVD discs :

1) 4:3 NTSC

2) 16:9 PAL

I wanna burn an unique DVD with both MPEG's.....

Choice for the target disc is : NTSC, 16X9;

So, one question borns : how can I convert a 4:3 NTSC file to 16:9 also NTSC ?
And if I were to do the opposite ?

The 4:3 NTSC file had been recorded on stand-alone recorder from TV, so I suppose it's interlaced, right ?

Any AVSynth script ?

Thanks,

Zetti.

P.S. Please forgive me other people that might have seen this question at another topic, the correct should be to start a new one;

[FulciLives]

I found a VERY simple way to deal with INTERLACED resizing ...

Go here to this website:

http://www.avisynth.org/stickboy/

Download the file called ---> jdl-util.avsi (first one on the list)

As per the directions put this in your AviSynth PLUGINS folder.

Now you say the NTSC MPEG-2 is from a RIPPED DVD that was created on a stand alone DVD recorder. So run the VOB files through DVD2AVI to get your D2V project file.

Your script should look like this:

LoadPlugin("mpeg2dec3.dll")
mpeg2source("D:\DVDRIP\VIDEO_TS\movie.d2v")
crop(0,60,-0,-60)
jdl_UnfoldFieldsVertical(flip=true)
LanczosResize(704,480)
jdl_FoldFieldsVertical(flip=true)
AddBorders(8,0,8,0)

OK what does this do exactly?

The whole jdl_UnfoldFieldsVertical stuff helps as "deal" with interlaced video. You have to be very carefull how you resize interlaced video as well as how you apply filters to it.

The jdl_UnforldFieldsVertical function takes care of ALL that.

So first we load our D2V project file. Then we crop 60 from the top and bottom. This is done to make it 16x9 anamorphic (after resizing it).

Then we have our jdl_FoldFieldsVertical function (start)

Now we resize the video (which has been cropped from 704x480 to 704x360) to 704x480. This crop then resize is what converts the video from 4:3 to 16x9

Now we have our jdl_FoldFieldsVertical function (end)

Last but not least I add 8 pixels to the right and left to bring the file from 704x480 to 720x480

Notice I am ASSUMING in this example that the source is 704x480

Why? I know some stand alone DVD recorders (such as the Panasonic brand) record 704x480 instead of 720x480

You could leave it 704x480 if you want by dropping the AddBorders command. Also if your source is 720x480 then you change the resize to read 720x480 (instead of 704x480 per my example) and drop the AddBorders command.

What if you want to use a filter?

Then you could try this:

LoadPlugin("mpeg2dec3.dll")
LoadPlugin("Convolution3d.dll")
mpeg2source("D:\DVDRIP\VIDEO_TS\movie.d2v")
crop(0,60,-0,-60)
jdl_UnfoldFieldsVertical(flip=true)
Convolution3D (0, 6, 10, 6, 8, 2.8, 0)
LanczosResize(704,480)
jdl_FoldFieldsVertical(flip=true)
AddBorders(8,0,8,0)

This is almost the same as the first script except I added the two lines for Convolution3D (with the values that equal the "movieLQ" mode).

*** WARNING ***

I just discovered this very fine piece of code (i.e., jdl_UnfoldFieldsVertical) but so far I've only used it once on a source that needed Convolution3D but NO resizing.

This script should work A-OK but you might want to run just a few minutes of your source and test it.

- John "FulciLives" Coleman

P.S.
For CCE use add this to the script (last line):
ConvertToYUY2(interlaced=true)

For TMPGEnc use add this to the script (last line):
ConvertToRGB(interlaced=true)

This info is for AviSynth 2.5x not the older 2.x version(s)

[Zetti]

Thanks Fulci, it's too late now in Brazil and I have to go to bed, I'll try tomorrow and return to you;

Thanks again,

Zetti

[Zetti]

Fulci,

I downloaded the TXT file, that's a source code in C....

Well, I copied it onto the plugins folder of v 2.54.....TMPGEnc says that this function doesn't exist......

I guessed in advance that wouldn't work........I see no way how AVSynth should access that source code..... any idea ?

Thanks so much, as always;

Zetti

P.S. Any Smooth deinterlacer for 2.54 ? Does it come built-in ?

[FulciLives]

Well ...

When you downloaded the file it is called this: jdl-util.avsi.txt

You have to rename it to this: jdl-util.avsi

I then simply put the file in my AviSynth Plugin folder which is located here on my computer:

C:\Program Files\AviSynth 2.5\plugins\jdl-util.avsi

Works for me!

- John "FulciLives" Coleman

P.S.
Go to AviSynth.org and click on external filters. It takes you to a place where you can download filters ... many exist both for 2.x and 2.5x

I think there is a version of Xesdeeni's SmoothDeinterlacer for 2.5x but I tried it once (a long time ago) and couldn't get it to work. I didn't know AviSynth well then so maybe the problem was me. I still use the version that works with 2.08 and just install and re-install between 2.08 and 2.54 as needed. I only use 2.08 with interlaced PAL to NTSC but with all other projects I use 2.54

[Zetti]

Well ...

When you downloaded the file it is called this: jdl-util.avsi.txt

You have to rename it to this: jdl-util.avsi

I then simply put the file in my AviSynth Plugin folder which is located here on my computer:

C:\Program Files\AviSynth 2.5\plugins\jdl-util.avsi

Works for me!

- John "FulciLives" Coleman

P.S.
Go to AviSynth.org and click on external filters. It takes you to a place where you can download filters ... many exist both for 2.x and 2.5x

I think there is a version of Xesdeeni's SmoothDeinterlacer for 2.5x but I tried it once (a long time ago) and couldn't get it to work. I didn't know AviSynth well then so maybe the problem was me. I still use the version that works with 2.08 and just install and re-install between 2.08 and 2.54 as needed. I only use 2.08 with interlaced PAL to NTSC but with all other projects I use 2.54

Thank you... ..I don't know C programing...

I'll be traveling tomorrow and will be back on Monday, then I'll return to it and contact you again, thanks so much;

I am involved in both conversions, PAL to NTSC at 16:9 and 4:3 to 16:9 NTSC;

Zetti

P.S. I searched all the night looking for SmoothDeinterlacer for v2.54, haven't found

It's boring to change between 2.54 and 2.08 all the time; also, it looks like MPEG2DEC3 is more advanced than the 2.08 version

[FulciLives]

I captured a 20 minute segment or so from my BLADE RUNNER Criterion NTSC LaserDisc.

This has an aspect ratio of 2.35:1 but of course is 4:3 so I am now testing that script out using Convolution3D and the crop/resize method to make it 16x9

This is a good test because I've been meaning to do this LD to DVD conversion anyways because the Criterion LD of BLADE RUNNER is the best cut of the film.

So if my sample goes well I'll just do the whole thing.

I'm also testing using iuVCR instead of my normal cap program so this test is also (for me) to see if the A/V sync is good.

I'll let you know how it turns out. It is encoding now ... in CCE from the captured AVI using AviSynth of course

- John "FulciLives" Coleman

P.S.
I'm using the AviSynth script I used in this thread converted for AVI source file instead of D2V.

[DJRumpy]

I tried iuVCR, but I had sync problems with it, where I've never had them with VirtualDub. I'd be curious as to whether or not it has improved any over the last two years.

[FulciLives]

Well my BLADE RUNNER clip looks very good

Here is the script I used:

LoadPlugin("Convolution3d.dll")
avisource("D:\CAPTURE\brtest_67ms.avi")
Trim(195,42620)
crop(0,60,-0,-60)
jdl_UnfoldFieldsVertical(true)
Convolution3D(0, 6, 10, 6, 8, 2.8, 0)
LanczosResize(688,480)
jdl_FoldFieldsVertical(true)
AddBorders(16,0,16,0)
ConvertToYUY2(interlaced=true)

The resize from 4:3 widescreen to 16x9 widescreen worked rather well with no problems what-so-ever. I used Convolution3D "movieLQ" mode.

So yeah this jdl thing works. In short it makes it very easy to process interlaced video

BTW the clip was captured at 688x480 because I just recently reverted back to the Avermedia drivers for my card and that is the resolution you have to use to get a proper aspect ratio.

As for using iuVCR ...

I'm using a new technique I discovered on the forums at the doom9 website

My capture was in perfect sync but I only captured the first 20 some odd minutes of the movie so I'll have to see what happens with longer captures.

Basically I set up iuVCR like this:



Then after the capture I used this program I found which "fixes" the stream offset (in this case it was -67ms).

That program is called AVI Offset Calculator and can be found here:
http://darkfalz.kiczek.com/software/

I read about that program here:
http://forum.doom9.org/showthread.php?s=&threadid=73901

I set up iuVCR the way I did per the suggestion of the person that wrote AVI Offset Calculator.

Anyways this method still needs to be tested more ... I have a funny feeling that it might not work for longer captures. Also if you have any frame drops (which I did not in my one test capture) I would imagine that this method might not be so great. However with most VHS sources I tend to get very little frame droppage (usually no more than maybe 4 to 8 frames through an entire movie) so maybe that's not enough to screw things up?

Anyways right now I'm doing a VERY long encode of a PAL to NTSC project so I won't be able to test any capture stuff til probably sometime early next week. This encode was started last night and won't finish til Sunday ... damn slow ass computer I have hehehe

- John "FulciLives" Coleman

[DJRumpy]

I'll stick with VirtualDub.

One program. No sync loss. Err..make that two programs. I use a scheduler program called 'VirutalDub Start Timer', which starts and stops the record, but I've been using it for two years without a single loss of sync. No editing or stretching required after the fact.

Fulci, I see an option there to 'Sync Using Stream Offset'. Wouldn't that be the equivelent of the VDub setting to sync audio to video? It might eliminiate any sync problems.

[FulciLives]

I'll stick with VirtualDub.

One program. No sync loss. Err..make that two programs. I use a scheduler program called 'VirutalDub Start Timer', which starts and stops the record, but I've been using it for two years without a single loss of sync. No editing or stretching required after the fact.

Fulci, I see an option there to 'Sync Using Stream Offset'. Wouldn't that be the equivelent of the VDub setting to sync audio to video? It might eliminiate any sync problems.

Well I have been using TheFlyDS to capture for a while now and it has been working well for me in maintaining audio sync but just recently I captured something that was about 2 hours long and I only had 6 dropped frames in that whole period of time (2 at a time spread out) and it was ever-so-slightly out-of-sync. Don't know why ... doubt it was the dropped frames.

Anyways with TheFlyDS I was using MASTER STREAM = AUDIO but that gives you an "odd" frame rate in the end ... usually something like 29.967 or 29.966 in my experiences.

Everyone seems to RAVE about iuVCR so I thought I'd give it a try again.

I've tried the audio resample trick in VirtualVCR and that didn't work for me. I also tried the audio resample trick with VirtualDubSync (I think that's the name of that mod) and again it doesn't work plus I take too much of a performance hit using Vdub anyways because of the VFW/WDM wrapper thing ... forces me to capture at 352x480 or I drop frames whereas with TheFlyDS/iuVCR/VirtualVCR I can do Full D1 resolution captures.

Anyways I'll be playing around some more with iuVCR ...

- John "FulciLives" Coleman

[Zetti]

Fulci and DJRumpy,

I am back home now from a quick trip and will try the method;

Fulci, what's the convolution filter intended for ? I just wanna make the 4:3 to 16:9 conversion...should I use it ?

Thanks,

Zetti

[Zetti]

Fulci,

This is my script for 4:3 to 16:9 NTSC conversion :

LoadPlugin("MPEG2DEC3.dll")
LoadPlugin("Convolution3d.dll")
mpeg2source("D:\Chris_Rea_Montreux_2002\TV\TV2.d2v ")
Trim(195,42620)
crop(0,60,-0,-60)
jdl_UnfoldFieldsVertical(true)
Convolution3D(0, 6, 10, 6, 8, 2.8, 0)
LanczosResize(704,480)
jdl_FoldFieldsVertical(true)
AddBorders(16,0,16,0)
ConvertToYUY2(interlaced=true)


TMPGEnc says :

"Convolution3D" supports YUY2 color format only ("path file", line 7);

I guess my source file is RGB (?).......any idea ?

Please note that I wanna make a strictly 4:3 to 16:9 conversion, I am not aware about the convolution function...

Thanks,

Zetti

[BSR]

Convolution3D is a noise filter that does both temporal and spatial filtering.
If you don't want any noise filtering, then leave it out.
Adding the filter does help compressability though.

[DJRumpy]

Zetti, the error just means that the filter requires a YUY2 colorspace. To convert it to YUY2, just add this line after your MPEG2SOURCE line:

ConvertToYUY2()

There is no space between the parens ().

The Convolution3D filter is usefull for analog captures. It helps to remove 'snow' and various other kinds of common noise. Try it with, and without, and see which you prefer.

Fulci, definately let me know if the sync setting in iuVCR helps. I'm not opposed to trying something new! I don't want to mess with post editing of the capture just to fix sync though. Let me know if that sync setting works or not. If it does, I'll give it a try again.

[Zetti]

Thanks BSR and DJ,

I'll try tonight again,

Zetti

[DJRumpy]

Just a note. I'm not sure what your encoding in, but if your encoder has a simular complaint, something like: RGB Required or somesuch, just add the following in at the end of yoru script:

ConvertToRGB()

Only add this if necessary, as it slows down processing. Most encoders accept YUY2 for input, so you should be fine.

[TerryG]

If you stick a 4x3 video in MemoriesOnTV.com and have MOTV set to do 16x9 it will put black boarders on the sides and keep the correct aspect.

[FulciLives]

Fulci,

This is my script for 4:3 to 16:9 NTSC conversion :

LoadPlugin("MPEG2DEC3.dll")
LoadPlugin("Convolution3d.dll")
mpeg2source("D:\Chris_Rea_Montreux_2002\TV\TV2.d2v ")
Trim(195,42620)
crop(0,60,-0,-60)
jdl_UnfoldFieldsVertical(true)
Convolution3D(0, 6, 10, 6, 8, 2.8, 0)
LanczosResize(704,480)
jdl_FoldFieldsVertical(true)
AddBorders(16,0,16,0)
ConvertToYUY2(interlaced=true)


TMPGEnc says :

"Convolution3D" supports YUY2 color format only ("path file", line 7);

I guess my source file is RGB (?).......any idea ?

Please note that I wanna make a strictly 4:3 to 16:9 conversion, I am not aware about the convolution function...

Thanks,

Zetti

Hello

Try this script:

Without Convolution3D this will work:

LoadPlugin("MPEG2DEC3.dll")
mpeg2source("D:\Chris_Rea_Montreux_2002\TV\TV2.d2v")
Trim(195,42620) 
crop(0,60,-0,-60) 
jdl_UnfoldFieldsVertical(true)  
LanczosResize(704,480) 
jdl_FoldFieldsVertical(true) 
AddBorders(8,0,8,0) 
ConvertToYUY2(interlaced=true)

Using Convolution3D this should now work:

LoadPlugin("MPEG2DEC3.dll")
LoadPlugin("Convolution3d.dll") 
mpeg2source("D:\Chris_Rea_Montreux_2002\TV\TV2.d2v")
Trim(195,42620) 
crop(0,60,-0,-60) 
ConvertToYUY2(interlaced=true)
jdl_UnfoldFieldsVertical(true) 
Convolution3D(0, 6, 10, 6, 8, 2.8, 0) 
LanczosResize(704,480) 
jdl_FoldFieldsVertical(true) 
AddBorders(8,0,8,0)

This setting for Convolution3D is called "movieLQ" which is a fairly light "video noise" setting intended for "very noisy DVD sources". There is an even "lower" setting (does less fltering" called "movieHQ" which is for "good DVD sources" and that setting is:

Convolution3D(0, 3, 4, 3, 4, 2.8, 0)

I usually use the "movieLQ" setting myself.

You should have gotten a file called, "Convolution3D.txt" with the Convolution3D download file (which is ZIP format I think) and that txt file has all of the settings listed.

ALSO PLEASE NOTE:
Your script looks like it is set up for a source that is 704x480 but you added 16 to each side. That would make it 736x480 so if you look at my two revised scripts you will see that I changed it to 8 on either side which will make it 720x480 as it should be

- John "FulciLives" Coleman

*** EDIT ***
I just realized that you said you were using TMPGEnc to do this encoding. If that is so then you need to do this:

If you use my revised script number one then change the last line to read like this:
ConvertToRGB(interlaced=true)

If you use my revised script number two then leave it as is BUT ADD this as the last line:
ConvertToRGB(interlaced=true)

Since your source is a DVD from a stand alone DVD recorder it is INTERLACED hense the (interlaced=true) but if the source is progressive then you would just use () like what DJRumpy said.

[Zetti]

Hi Fulci and all,

Well, yesterday night I fixed my script and loaded it into TMPGEnc, initially, I found a very strange error : it skipped some seconds of the beggining and started from say the 3rd or 4th second....

Then, although the MPEG2 file is 45 min long, it encoded just 14 minutes and ended with no error message.....I ain't at home now so I can't copy/paste my script now, but I believe it's correct (I used the light convolution filter) and got strange mistakes,

Well, any guess ?

Thanks,

Zetti

[DJRumpy]

Try remarking out the convolution bits of the script by placing a # sign in front of those lines. See if the problem persists.

[Zetti]

Try remarking out the convolution bits of the script by placing a # sign in front of those lines. See if the problem persists.

DJ,

Looks like you've got it ! This is my script now :

LoadPlugin("MPEG2DEC3.dll")
LoadPlugin("Convolution3d.dll")
mpeg2source("D:\Chris_Rea_Montreux_2002\TV\TV2.d2v ")
#Trim(195,42620)
#crop(0,60,-0,-60)
#ConvertToYUY2(interlaced=true)
#jdl_UnfoldFieldsVertical(true)
#Convolution3D(0, 3, 4, 3, 4, 2.8, 0)
#LanczosResize(704,480)
#jdl_FoldFieldsVertical(true)
#AddBorders(8,0,8,0)
#ConvertToRGB(interlaced=true)

At least now I've loaded it into TMPGEnc and the very first frame appears properly, started encoding now, let's see how it goes when it finishes tomorrow...

Thanks,

Zetti

P.S. BTW, thanks for the IFOEdit PAL / NTSC patch guide

[DJRumpy]

You remarked out a bit more than I would have

I just re-read your script. Try remarking out ONLY the TRIM command. The TRIM command is telling AVISynth to start at the 195 frame of your video, and stopping at the 42,620th frame. I assumed you had it there for a reason, but I'm guessing since your reporting this problem, you don't. I guess someone pasted into an example, and you kept it by mistake.

Unless you need the video between those specific frame numbers, you should remove it. That command is usefull for removing things like commercials and such, but you typically don't use it on source files from DVD projects.

Trim(195,42620)

Make sure you play your .AVS script in Windows Media Player before encoding. It will play just like an AVI does. You can verify the video length that way, and ensure it looks right. You can just associate your .AVS file with Windows Media Player, or just drag and drop the .AVS file onto WMPlayer. Either way, you should always preview it to make sure it looks right. The only gotcha is that some noise filters are very cpu intensive, and if your PC is too slow, it may not play smoothly. Not to worry though, as it will encode just fine.

[Zetti]

Thanks DJ,

I have stopped it quite in time, as far as I realized, the # deletes the line

I copied/pasted Fulci's script, hence the mistake on mine, now I have previewed through Media Player, it seems correct;

I'll see how it ends tomorrow morning;

Thanks,

Zetti

[DJRumpy]

In essence, it does. It remarks out the line, so it's completely ignored by AVISynth. You can put the # sign anywhere in a line. Anything after it is ignored completely.

I'm glad you caught it. Nothing is more frustrating than waiting on TMPGenc to finish, only to find it's not right.

Sorry I didn't catch that before.

[FulciLives]

Hi

Sorry for being late here I was away last couple days.

My fault. I had that TRIM command in my last post. It was from a script I was actually using and was modifying it for you and I forget to get rid of it. That is why your file was starting several seconds in and ending short.

So the scripts I posted should work as they are ... just delete that trim line (or insert your own if you need to edit it at all).

- John "FulciLives" Coleman

P.S.
You can load your file into VirtualDubMod and edit it then go to TOOLS and SCRIPT EDITOR and a window pops up called VIRTUALDUBMOD SCRIPT EDITOR. There is an option called, "IMPORT FRAMESET AS TRIMS" and doing this will translate your edits into a "trim" line for AviSynth.

[Zetti]

Dear DJ and Fulci,

As always, thanks for your attention

Well, I don't have good news, this is my actual script :

LoadPlugin("MPEG2DEC3.dll")
LoadPlugin("Convolution3d.dll")
mpeg2source("D:\Chris_Rea_Montreux_2002\TV\TV2.d2v ")
crop(0,60,-0,-60)
ConvertToYUY2(interlaced=true)
jdl_UnfoldFieldsVertical(true)
Convolution3D(0, 3, 4, 3, 4, 2.8, 0)
LanczosResize(704,480)
jdl_FoldFieldsVertical(true)
AddBorders(8,0,8,0)
ConvertToRGB(interlaced=true)

Results : I can't see the head of the guitar player anymore

Neither the song titles at the bottom of the screen....

It seems it has increased the vertical size of the picture, and now it doesn't fit anymore at my regular 4X3 TV screen....

Any guess ?

Thanks for helping, Fulci, my fault on not realizing the "trim" command, at least if I had read it more deeply, I would have guessed that...

Thanks,

Zetti

P.S. I always have issues also on my PAL to NTSC conversion following Xesdeeni's method

Remember Gordian Knot indicated the file (another one) as being progressive, while TMPGEnc indicated it as being field A first ?
I encoded it through the 3 AVIsynth scripts : progressive, field A and field B.
Field B = horrible, it's not correct
Progressive = interlaced artefacts
Field A = less worse, still, not smoothy video on fast scenes

[FulciLives]

Well remember you are attempting to convert 4:3 material to 16x9 and the only way to do that is to cut 60 pixels off of the top and 60 pixels off of the bottom.

So yes you are going to loose picture information UNLESS the original 4:3 is widescreen (with an aspect ratio of at least 1.78:1 or "wider" such as 2.35:1 etc.). There are many DVD discs that are widescreen but in a 4:3 ratio rather than a 16x9 ratio. I assumed since you were attempting to convert 4:3 to 16x9 that you had such a case.

Now it sounds like your 4:3 video is full screen (fills the screen from top to bottom) so yes you will loose picture information by converting it to 16x9

So you should leave it at 4:3

Now there is a way to make 4:3 full screen 16x9 anamorphic but it will only look good on a true 16x9 TV screen ... it will look rather "odd" on a 4:3 TV screen as the 4:3 full screen image will have black on all 4 sides of the image when viewed on a 4:3 TV.

I wouldn't do it myself unless:

A.) You are mixing 4:3 full screen and 16x9 on one DVD and want the whole thing to be 16x9

B.) It will only ever be watched on a true 16x9 widescreen TV screen.

DJRumpy did an article once on aspect ratio and I probably should read it now to refresh myself (it's been a while) but if I recall correctly a 16x9 DVD gets stretched out on playback from 720x480/576 to 960x480/576

Maybe DJRumpy can verify that number and if it is wrong give us the correct number but assuming 960 is correct then you would do this:

NTSC 4:3 FULL FRAME TO 16x9 WS

LoadPlugin("MPEG2DEC3.dll") 
LoadPlugin("Convolution3d.dll") 
mpeg2source("D:\Chris_Rea_Montreux_2002\TV\TV2.d2v")  
ConvertToYUY2(interlaced=true) 
jdl_UnfoldFieldsVertical(true) 
Convolution3D(0, 3, 4, 3, 4, 2.8, 0) 
jdl_FoldFieldsVertical(true) 
AddBorders(128,0,128,0)
LanczosResize(720,480) 
ConvertToRGB(interlaced=true)

OK so what does this do?

First we padd up the original image (which we know is 704 width) to a width of 960 by adding 128 pixels of black on either side of the screen. Then we resize this 960x480 to 720x480 which makes it 16x9 anamorphic.

Now on a 16x9 TV the image will fill the screen from top to bottom (without any image loss) but will have black on either side of it (with the image centered) but on a 4:3 TV the image will be centered but will have black on all 4 sides of the image.

- John "FulciLives" Coleman

[Zetti]

Thanks Fulci, yes, I understand...

I'll try the new script anyway tomorrow and see how it goes (PC is busy now); I have nothing to lose but time (and eletric power) :P

Thanks to DJ as well, I'll return here as soon as I have results;

Zetti

[DJRumpy]

Just a suggestion. You can do a less drastric aspect if your source is fullscreen 4:3.

Instead of making a 'proper' 16:9 anamorhpic widescreen conversion, you can sort of meet half way, using a less drastic aspect ratio (something closer to 4:3 ). Instead of cutting off 120 total pixels (2.35:1), consider trying for a 1.77:1 conversion.

Your crop command would look like this:

Crop( 0,48,0,-48 )

This will create a true widescreen 16:9 video (non-anamorphic), converting your 1.33 (4:3) video to 1.77, instead of 2.35. The change in aspect is a bit less drastic, cutting off a bit less of your source. I too assumed your source was letterboxed 4:3, and not fullscreen 4:3.

Another trick I've used, is to crop more from the bottom, than the top. Film makers seem to concentrate more on the upper two-thirds of the screen, so you could just crop more from the bottom. Just keep the total amount you crop consistant (48*2=96). You could crop 60 from the bottom, and that would leave only (96-60=36) 36 pixels from the top. In that case, your crop command would look like this:

Crop(0,36,0,-60)

The total amount cropped would still give you a 16:9 video (1.77 aspect).

And finally, you don't have to follow the rules. It's your video. You can always crop off less than the exact value required to make your video 1.77. I find that, with a widescreen TV, watching fullscreen video all the time in 'wide zoom' mode, you don't notice very much distortion when you have something in between 1.77 (16:9), and 1.33 (4:3). You can always crop less than 96 total pixels (try 80, or some othe value), and see if the playback appears too 'short'.

Just a suggestion...