Comments on the DV to DVD guide

[bb]

If you have any comments, additions, corrections, simplifications, improvements, hints, opinions, etc. on the new "DV to DVD guide", please put them in here.

bb

[RathO]

Hi bb, i only want to say that it is well done. Nice work!
This guide is a GOOD start for anyone who wants to do a dv->dvd conversion.

You don't show it as a "one-strict-way-process", so it gives ideas and push the user to experiment new avenues by him-self. ;)

Regards

[jikchung]

bb,
I also wanted to say nice work. Just got my first DVD burner, and am starting to transfer home video, this will be a tremendous help.
Thanks!

[bb]

Thank you very much for the flowers. Criticism will be appreciated, too.

bb

[FredThompson]

NTSC DV is 4:1:1 with the potential for significant problems. trbarry and Xesdeeni have created solutions for AviSynth and VirtualDub:

http://mywebpages.comcast.net/trbarry
http://www.geocities.com/xesdeeni2001

4:1:1 uses one color value for 4 adjacent horizontal pixels. DVD is 4:2:0 using one color value for a square of 4 pixels. The two formats don't play well together and you will find results are better is the chroma is corrected beforehand.

--

DV codecs have varying performance. DV camcorders typically work with a wider luma range than is considered "safe" for TVs. Take your pick, old-style TV "safe" or wider luma range. It's an ongoing debate which is "better."

My opinion is evident, I encode for now and the future, not what was.

Canopus codec appears to only work with the "safe" range.

Codec choice is a very subjective decision. While quite a few people make claims about the performance of various codecs, there doesn't appear to be any kind of comparison against a true "reference" codec. Keep in mind Panasonic codec and various...uh...modified codecs may be quite outdated. The point is, chose the codec for which you prefer the results.

--

Another debate is width of image. Camcorders record 720x480, most TV is broadcast as 704x780 with some black borders.

[Dali Lama]

Hi bb,

Good Job.

Here are my suggestions:

Include a segment on how to convert from DV type 1 to type 2. There is a free DV converter from canopus.com.

Try to illustrate that using Convolution3D on an interlaced source as you outlined will slow down encoding considerably and may justify spending more time on motion search. Case in point, if you are going to filter the source, spend the time.

Speaking of filtering please add that using those settings on Convolution3D smooth the image significantly and that encoding into MPEG-2 for playback on tv dramatically diminishes the common noise found in DV (such as low light conditions).

Since you mentioned ProCoder, why not just describe how to set it up? Its really simple I think. The defaults are very good for DV material.

Also, speaking of TMPGEnc, make sure to mention that using the popular Canopus DV codec, you SHOULD set "output YUV data as YCbCr" either in the "quantize matrix" tab or "global environmental settings" I don't think both are necessary. Using other DV Codecs don't require this feature, unless it keeps the DV range 16-235 intact.

Also, I think mentioning that you can select "interpolate YUV data from 4:1:1 to 4:4:4 in the same tab for the Canopus Codec, which as Fred said is important.

Finally, I don't think the use of DC precision 9 over 10 is clear. Do you have a reason why you use that? Most documents, including the ones for TMPGEnc and CCE say that DC precision 10 is better (slightly slower processing) especially in low light conditions (perfect for DV conditions).

Thanks again and I hope I could help a little,

Dali

[FredThompson]

Oh, yeah, I'd forgotten about near-black noise. There are two smoothers I know of that work in that range. Both are RGB, though. ShadowSmoother and Darksmoother are the names. You can read about them here: http://forum.doom9.org/showthread.php?s=&threadid=56517

That's a very good thing to do because DV and camcorder source is notorious for problems with low-light or restricted-light. I have problems with non-incandescent light like those long-life bulbs. There's must be something missing that human eyes aren't bothered by.

[bb]

Originally posted by FredThompson
DV camcorders typically work with a wider luma range than is considered "safe" for TVs. Take your pick, old-style TV "safe" or wider luma range. It's an ongoing debate which is "better."
As far as I understood it, if the luma range is 0-255, you could reduce it to 16-235 during MPEG-2 encoding. So why care about this at the DV level?

Another debate is width of image. Camcorders record 720x480, most TV is broadcast as 704x780 with some black borders.
Do you mean you'd get a wrong aspect ratio if you encode 720x480 DV to 720x480 MPEG-2?

bb

[bb]

Originally posted by Dali Lama
Include a segment on how to convert from DV type 1 to type 2.
Why? The guide works with either type-1 or type-2. Ok, if the user doesn't want to use DirectShowSource, then there's currently no way around the conversion (except for a frameserving method through VirtualDub I just read about).

Try to illustrate that using Convolution3D on an interlaced source as you outlined will slow down encoding considerably and may justify spending more time on motion search. Case in point, if you are going to filter the source, spend the time.
Yes, but I believe you still get better results if you get rid of as much noise as possible.

Speaking of filtering please add that using those settings on Convolution3D smooth the image significantly and that encoding into MPEG-2 for playback on tv dramatically diminishes the common noise found in DV (such as low light conditions).
Noise can dramatically reduce compressibility and therefore the overall picture quality at a given bitrate, too. If you don't use integrated filtering in the MPEG-2 encoders, I can't see why the noise would be diminished.

Since you mentioned ProCoder, why not just describe how to set it up? Its really simple I think. The defaults are very good for DV material.
Simple reason: I don't have Procoder. Would you like to write a small Procoder addendum?

Finally, I don't think the use of DC precision 9 over 10 is clear. Do you have a reason why you use that? Most documents, including the ones for TMPGEnc and CCE say that DC precision 10 is better (slightly slower processing) especially in low light conditions (perfect for DV conditions).
I'm open to changes in this one. I got the recommendation of 9 bits from various guides, e.g. the one from mb1. 10 bits is said to be exaggerated, but I never did extensive testing on this.

Thanks for the input.

bb

[bb]

Originally posted by FredThompson
Oh, yeah, I'd forgotten about near-black noise...
Good point. I think this is more for advanced users, though, not really needed for a basic guide. Maybe there should be an extension to the guide to point out some advanced topics.

bb

[FredThompson]

Originally posted by bb
As far as I understood it, if the luma range is 0-255, you could reduce it to 16-235 during MPEG-2 encoding. So why care about this at the DV level?I can think of no good reason to reduce to 16-235 unless that's the original range. If the source is true DV, reducing the range screws it up. On top of that, the 16-235 thing is based on old tube technology. It doesn't make sense to me to take proper DV of 0-255 and squish the luminence.Do you mean you'd get a wrong aspect ratio if you encode 720x480 DV to 720x480 MPEG-2?I have no idea why I typed that unless I was thinking about how DV camcorder is the full 720 and most other media is 704. I can't think of much of a practical use other than I, personally, crop then letterbox to center but that's just being anal.

[FredThompson]

Originally posted by bb
Good point. I think this is more for advanced users, though, not really needed for a basic guide. Maybe there should be an extension to the guide to point out some advanced topics.It does have an effect on compression, the same way noise will. I mentioned it primarily because it's a great way to squeeze some more performance without reducing important detail.

Another thought might be a proper comb filter if the source was analog tape passed through a DV converter.

Uh...yeah...you're right. Filtering quickly gets into the "advanced" arena.

[bairradino]

I assume that it is not possible to frameserve directly from Premiere6.5 to MPEG-2 encoder (without the need of duplicating HD space).
Is it true?

[Dali Lama]

Originally posted by bb
Why? The guide works with either type-1 or type-2. Ok, if the user doesn't want to use DirectShowSource, then there's currently no way around the conversion (except for a frameserving method through VirtualDub I just read about).

When I use DirectShowSource() I get incredibly slow frameserving, that's why I suggested it. Also, you can convert from type-1 to type-2 using the Canopus DV Converter (the reference AVI setting doesn't take up that much space)

Concerning Procoder, I can write a simple guide.

Also, encoding into MPEG-2 will reduce noise, since it also reduces detail. But I agree that it won't reduce a lot of noise and having noise/detail will reduce compressibility. However, don't forget TVs reduce noise too (built in filtering).

bb, the guides for CCE and TMPGEnc both say DC precision of 10 will produce better results, especially in dark areas. It slows down encoding, but not by much.

bye,

Dali

[FredThompson]

Originally posted by bairradino
I assume that it is not possible to frameserve directly from Premiere6.5 to MPEG-2 encoder (without the need of duplicating HD space).
Is it true? http://www.debugmode.com/pluginpac/frameserver.php
http://www.videotools.net/index.php?rub=guides&htm=intro2pvsp.htm

[SomeJoe]

Originally posted by FredThompson
I can think of no good reason to reduce to 16-235 unless that's the original range. If the source is true DV, reducing the range screws it up. On top of that, the 16-235 thing is based on old tube technology. It doesn't make sense to me to take proper DV of 0-255 and squish the luminence.

"Proper" DV stores blacks at a luminance value of 16 and whites at a luminance value of 235. All ITU-R 601 conformant digital video, including DV, D1, DVCPro, etc. are like this.

The difference between DV codecs comes in when the codec is used to decode the stored 5:1 compressed DV YUV data and hand off an RGB frame to the application that requested it. Most modern DV codecs, including the Canopus and Matrox codecs render a DV YUV pixel with a luminance value of 16 (pure black) into an RGB pixel value of 16,16,16, which is a dark gray on the computer monitor. Simlarly, a DV YUV pixel with a value of 235 is rendered into a 235,235,235 RGB pixel which is not quite fully white on the computer monitor.

This accounts for the 16-235 or 0-255 settings for most MPEG-2 encoders. They have to know what reference level your DV codec considers "black" and "white" when the DV frame has been rendered to an RGB frame. The 16-235 choice is correct for most modern DV codecs.

Some DV codecs give you the full RGB range of 0-255 when rendering their RGB frames, which looks better on the computer monitor. It ensures that a black DV pixel becomes a black computer pixel. However, this can cause a problem called the "white clip" problem.

The reason modern DV codecs render to RGB with the 16-235 range is that some cameras can capture DV YUV pixels that are outside the 16-235 range. For instance, if you pan a DV camera across a reflection of the sun in a window, some white luminance YUV values will be captured by the camera that are beyond 235 - aka "superwhite" values, due to ringing/overshoot in the signal captured by the CCD. This is perfectly fine as far as the camera is concerned - these values are legal in the DV YUV space. Where they're not legal is when they're converted to analog (they'll be greater than 100 IRE, since Y-value of 235 = 100 IRE).

In this situation, a problem will arise if you have a DV codec that renders to RGB using a 0-255 range rather than 16-235. If your DV codec renders a 235 luminance DV YUV pixel to RGB 255,255,255 (where it looks white on the computer monitor), then all the highlight detail in that superwhite shot will be lost, because any YUV pixel greater than 235 gets clipped to an RGB value of 255,255,255. Modern DV codecs that render a DV YUV 235 pixel to RGB 235,235,235 will have headroom left in the conversion to retain the superwhite values and keep the highlight detail.

In short, you don't have to do anything to the DV's luminance range with AVISynth as long as you tell your MPEG encoder what range your DV codec renders RGB frames to (16-235 or 0-255). For optimal conversion without white clip, specifically using a DV codec than renders RGB frames at 16-235 will solve that.

For more info, see Adam Wilt's white clip (http://www.adamwilt.com/DV-FAQ-editing.html#WhiteClip) page.

[DDogg]

bb, you mentioned you have not included ProCoder because you do not have it, but I am sure there is a demo for ProCoder that you could get easily and with very minimal effort. Since ProCoder is arguably superior for DV/interlaced conversion to SVCD/DVD, and is a simple drag and drop operation, the exclusion is puzzling in a guide titled DV to DVD conversions.

OT, btw, great work on the guide and for bringing the DV-T1 problem to Sh0dan's attention. We will all benefit from you taking the initiative on solving this vexing problem.

[FredThompson]

Originally posted by SomeJoe
In this situation, a problem will arise if you have a DV codec that renders to RGB using a 0-255 range rather than 16-235. If your DV codec renders a 235 luminance DV YUV pixel to RGB 255,255,255 (where it looks white on the computer monitor), then all the highlight detail in that superwhite shot will be lost, because any YUV pixel greater than 235 gets clipped to an RGB value of 255,255,255. Modern DV codecs that render a DV YUV 235 pixel to RGB 235,235,235 will have headroom left in the conversion to retain the superwhite values and keep the highlight detail.I disagree. Go to http://www.geocities.com/fredthompson6/ and follow the link called, "Canopus DV Codec, MainConcept DV codec, broken YUV conversion". The leftmost image is Canopus freebie decoding to 16-235, the middle is MainConcept 2.3.1 decoding to 0-255. That's magnified 400%. Where is the loss in the over-saturated white indicated by the pink arrow?

Forthermore, suppose we run with your statement that some DV camcorders do 0-255 and some do 16-235. How do you propose to determine the range of source? If a camcorder shoots 0-255 (if we think in RGB terms) coring or squishing that to 16-235 alters the original range. Just because there is "THE standard" which states a more limited luminence range should be used does not mean I MUST use it nor is it the final authority.

I WILL grant you that near black and near-white typically have very little useful information...typically, not always.

ITU-R 601 is a relatively old standard, is it not? Was it created with plasma or computer monitors in mind? How about 1080i?

FWIW, Adam makes it very clear he does not like the DV format. While he has good information, he is quite biased against DV and some of his reasons make a tremendous amount of sense. (If you want to research that, I think it was on one of the Mac-oriented boards, maybe one that has columns he writes.) Nonetheless, the idea that ITU-R 601 must be strictly followed ignores advances in technology and can have the effect of unneccesarily restricting quality of final product.

[bb]

Originally posted by DDogg
bb, you mentioned you have not included ProCoder because you do not have it, but I am sure there is a demo for ProCoder that you could get easily and with very minimal effort. Since ProCoder is arguably superior for DV/interlaced conversion to SVCD/DVD, and is a simple drag and drop operation, the exclusion is puzzling in a guide titled DV to DVD conversions.
Well, I'm glad I found the time to set up at least some guide, so I had to focus on what I had already available and was familiar with. I'll be glad to add a Procoder guide as soon as possible, whenever I get the time, or Dali Lama sends me his guide.

For the same reason there Mainconcept's MPEG-2 encoder has not been mentioned, and there may be one or two others being worth to be explained.

OT, btw, great work on the guide and for bringing the DV-T1 problem to Sh0dan's attention. We will all benefit from you taking the initiative on solving this vexing problem.
See http://forum.doom9.org/showthread.php?s=&threadid=60766

bb

[bb]

Originally posted by Dali Lama
When I use DirectShowSource() I get incredibly slow frameserving, that's why I suggested it. Also, you can convert from type-1 to type-2 using the Canopus DV Converter (the reference AVI setting doesn't take up that much space)
There's a new AviSynth version available (see post above).

Concerning Procoder, I can write a simple guide.
Thank you very much. I guess DDogg can't wait to see it :D

bb, the guides for CCE and TMPGEnc both say DC precision of 10 will produce better results, especially in dark areas. It slows down encoding, but not by much.
Higher precision consumes more bitrate, but improves quality. Thus at a given bitrate there is break-even on best quality regarding DC component precision vs. bitrate. I'm not sure whether 9 bits or 10 bits will result in best overall quality for DVD, but I know that you shouldn't use more than 8 bits for SVCD.

bb

[DDogg]

heh, well my guide for ProCoder would be:

1> Create droplets for svcd or dvd creation and put on desktop.
2> Drag your DVCam DV.avi on top of correct droplet and wait for a while.
3> Burn and watch great DV to DVD conversion that looks as good, or probably better, than any other.

I didn't figure that would quite fit in well within the complexities of special avisynth scripts, unusual filters, colorspace conversions, etc., and all this other high tech talk. I guess I am just dumber than mud :)

[FredThompson]

Dumber than mud? I doubt it. Try the 411->422 conversions on NTSC DV and look at things like logos, window pane dividers, etc. and you'll see the shimmering is greatly reduced. The extreme examples we've used from the Star Wars laserdisc of saturated red on a black background really shows the problem. It's worth the effort, believe me. PAL DV doesn't have these issues becuase it's 4:2:0, it's got other things to worry about.

[kempfand]

@ Fred: PAL DV ... it's got other things to worry about :confused:
Could you elaborate a bit more ? I live in PAL-world and would like to learn more

Kind regards,

Andreas

[FredThompson]

http://www.hometheaterhifi.com/volume_8_2/dvd-benchmark-special-report-chroma-bug-4-2001.html

I'm not sure what the other problems are. Ask I4004 on Donald Graft's forum. He can explain.

[Dali Lama]

[list=1]
Enter Source Tab -> Add DV File | If you have multiple sources use the "stitch" command and you can use the arrows to modify their placement
Enter Advanced Tab to modify "in" and "out" points or add filters
Enter Target Tab -> Choose "Add" -> Choose "VCD/DVD" -> "MPEG2 DVD (NTSC)" or "MPEG2 DVD (PAL)
Enter Convert Tab -> Choose Convert
[/list=1]

Yes, Using Procoder is very easy. But 99.9% of its users are going to want to edit the DVD Templates. Lets talk about that.

At step 3 you can enter "Advanced" to modify the DVD template and save it (if you wish).

Note: I won't talk about all of the options as most are self-explanatory. :)

[list=1]
Under "Interlacing" parameter you will see it is defaulted to "upper/top field first". All DV sources are "lower/bottom field first". (DDogg, I hope you knew that ProCoder defaulted to this!)
Under "Aspect Ratio" parameter select your aspect ratio. Most DV is 4:3, but if yours is 16:9 just select it. Or if you have cinemascope 2.35:1 then use that. Also, set "pixel aspect ratio" to desired ratio (note this is not the same as "video aspect ratio", just use the drop down menu).
Under "Speed/Quality" parameter simply select the highest quality your willing to wait for. All produce acceptable results at high bitrates.
Under "CBR/VBR type" parameter select CBR if you are using a high bitrate for DVD. Putting about 60 minutes DV footage on one 4.7GB DVD is HIGH BITRATE. 90 minutes is medium, 120 minutes is low bitrate. 2-Pass VBR should be used otherwise for hitting your target filesize and maintaining the highest possible quality. 1-Pass VBR for the same situation if you cannot wait for 2 processing cycles.
Set maximum bitrate to 9000 or 8000 kbps.
Under "Use Closed GOP" select it if you want to be able to seek faster in your movie (probably not needed). Using open GOP (unchecked) results in better compression.
Under "Distance Between I/P frame" automatic is fine, but if you use high bitrate then select 1 frame (for no B frames). If you are using low bitrate then select 3 frames (for 2 B frames).
Select "3/2 Pulldown" if your camcorder records in 24P!!!
Under "Picture Structure" select "Field Structure" for interlaced sources and "Frame Structure" for progressive sources. "Automatic Selection" is not compliant with all DVD players, but if it works it might benefit DV interlaced sources in low/no motion areas.
"DC Precision" IMO should be set to 10 unless your footage is mostly slow motion and composed of well lit scenes, then select 9.

Ok, that's it. If I made a mistake or forgot something, please contribute. I hope bb can salvage this guide and make it sensible :)

All the best,

Dali

[/list=1]

[bb]

First of all thank you very much for your contribution, Dali.

Now a few comments:

Originally posted by Dali Lama
Under "Use Closed GOP" select it if you want to be able to seek faster in your movie (probably not needed). Using open GOP (unchecked) results in better compression.
Usually "Closed GOP" is not used for DVDs. As you said not to close the GOPs gives better compression, so I'd simply recommend that.
Under "Distance Between I/P frame" automatic is fine, but if you use high bitrate then select 1 frame (for no B frames). If you are using low bitrate then select 3 frames (for 2 B frames).
I highly doubt that using a "no B-frame" option is a good idea, even for high bitrate encodings. Some people use "I-frame only" encodings, but that's for capturing purposes and the easy of editing. I would always use I-, P-, and B-frames.

bb

[FredThompson]

eyup, what he said

wrt closed GOPs, that tends to make things a lot blockier because it ups the bandwidth requirement.

[jfcarbel]

I do alot of analog captures from older VHS tapes using my ADVC-100 to capture it to DV.

Does this guide still apply to DV that was created from analog captures or is it only for straight DV from a Camcorder?

For VHS captures I would think the Convolution3D settings would need to be changed. Can someone suggest some settings or maybe a sticky guide can be done for analog DV VHS captures.

Thanks for the Guide!

[bb]

jfcarbel,

you can use the guide for your purpose. You'll probably need stronger filtering; you can use e.g.
Convolution3D(preset="vhsBQ")
This has been designed for VHS sources, but of course the best way is to experiment.

bb

[SomeJoe]

Originally posted by FredThompson
I disagree. Go to http://www.geocities.com/fredthompson6/ and follow the link called, "Canopus DV Codec, MainConcept DV codec, broken YUV conversion". The leftmost image is Canopus freebie decoding to 16-235, the middle is MainConcept 2.3.1 decoding to 0-255. That's magnified 400%. Where is the loss in the over-saturated white indicated by the pink arrow?

Well, there's several things to consider here. How do we know the Main Concept codec is rendering to RGB 0-255 or that the Canopus codec is rendering to RGB 16-235? (I have no idea which way either one renders, nor am I saying you're wrong - I just would like to know the test method that determined this).

Also, there is the question of whether the white is over-saturated in the original DV. Generally, you would only get over-saturated whites (i.e. DV YUV pixels with >235 luma) in video that was shot on a DV camcorder. If the source is from something else and was captured with a DV codec or converted to DV, you probably won't have any over-saturated whites in the frames.

Originally posted by FredThompson
Forthermore, suppose we run with your statement that some DV camcorders do 0-255 and some do 16-235.

I never said that. To my knowledge, all DV camcorders record according to ITU-R 601, using 16-235 luma range. The operation that I'm referring to that can cause problems is the DV codec used in the PC, and how it renders the YUV-encoded DV frames to RGB.

Originally posted by FredThompson
How do you propose to determine the range of source? If a camcorder shoots 0-255 (if we think in RGB terms) coring or squishing that to 16-235 alters the original range.

Source range shouldn't be an issue, since the camcorder should be ITU-R 601 compliant. The codec used on the PC is the problem. Hopefully, your answer to my test methodology question above should answer how we should go about determining a codec's RGB rendering characteristics.

Converting video that uses 0-255 as a luma range to 16-235 does indeed reduce the accuracy of how the luma values are described (it's the equivalent to increasing the quantization error in digital audio by reducing the number of bits per sample). But it doesn't necessarily alter shadows/highlights unless there is a disagreement between the references of the original and the converted video. i.e. as long as we agree that in the 0-255 video, 0=black and 255=white, and in the converted video 16=black and 235=white, then the resulting converted video should have no brightness shift. It will just have (slightly) coarser luma gradients.

Originally posted by FredThompson
Just because there is "THE standard" which states a more limited luminence range should be used does not mean I MUST use it nor is it the final authority.

Agreed, but interoperability between different manufacturers' products is enhanced when they all follow the same standard.

ITU-R 601 has been around a while, but because of it's prevelance and dominance, most equipment is built around it. Since the equipment that we use (camcorders, decks, codecs, etc.) is centered around that standard it becomes convenient to work within it. No one is saying that you yourself should concede to the standard if you don't want to - certainly there are improvements that can be attained. But if you want to shoot your own video with a camcorder, I believe you will be hard-pressed to find one that records 0-255 on the tape.

[bb]

Originally posted by FredThompson
NTSC DV is 4:1:1 with the potential for significant problems. trbarry and Xesdeeni have created solutions for AviSynth and VirtualDub:

http://mywebpages.comcast.net/trbarry
http://www.geocities.com/xesdeeni2001

4:1:1 uses one color value for 4 adjacent horizontal pixels. DVD is 4:2:0 using one color value for a square of 4 pixels. The two formats don't play well together and you will find results are better is the chroma is corrected beforehand.

Hi FredThompson,

would you be so kind to post a small AVS snippet to illustrate the use of ReInterpolate411? I can't test it, because I don't have NTSC sources.

bb

[bb]

Are there any comments on the 16:9 part? Did anyone notice that I didn't talk about anamorphic encoding? (I'll add that as soon as I get the time).

bb

[f@chance]

@ Dali Lama

I just captured from a Laser Disc the SW Definitive edition in PAL and would like to transfer it to DVD. I have searched the forum and read the DV to DVD guide and came across your quote. I am using Canopus Procoder for converting the DV file to MPEG2 and I am using the DVD/PAL template.

Under "Aspect Ratio" parameter select your aspect ratio. Most DV is 4:3, but if yours is 16:9 just select it. Or if you have cinemascope 2.35:1 then use that. Also, set "pixel aspect ratio" to desired ratio (note this is not the same as "video aspect ratio", just use the drop down menu).

How can I find out what the AR is of my capture and what should the setting be for a WS TV? The display will be eventually when the money is there, a progressive plasma display.

"DC Precision" IMO should be set to 10 unless your footage is mostly slow motion and composed of well lit scenes, then select 9.

What is DC precision? The Procoder Help file has no entry for it and the default setting is 8.

Finally since I have a progressive DVD player for playing back PAL progressive streams, will it need the Adaptive Interlacing filter?

Thanks for any help.

F@Chance

[FredThompson]

Originally posted by bb
Hi FredThompson,

would you be so kind to post a small AVS snippet to illustrate the use of ReInterpolate411? I can't test it, because I don't have NTSC sources.

bb
Reinterpolate411()

Both sites have magnified images that show the effect. Xesdeeni's is better because it compares different approaches.

I'll certainly redo my page soon. Right now I've got 4 of 5 computers down so anything like that isn't going to happen soon. Some a** hacked my business email/fax/voice computers.

[bb]

Canopus ProCoder guide added.

bb

[insertdisk]

I was just reading in the DVD authoring section and stumbled across a thread that said BeSweet could not be relied on to make a DVD compliant Ac3 stream. Has anyone experienced this? I would like for my DVDs to be playable in most any player and wonder if this is a real problem or has been fixed since that post. See the post here
http://forum.doom9.org/showthread.php?s=&threadid=60510

[troy]

I tried making mp2 files for my dvds using beesweet. Dvdlab would not accept them. For all the praise I read about beesweet I was definately dissapointed. I used the mp2 for dvd matrix in beesweet. I then used the very crude toolame gui and toolame. Presto dvdlab accepts these good. What is with all the praise people have for beesweet when it cant even produce a compliant mp2 stream.

[FredThompson]

user error

[troy]

ok maybe it is user error. what do I need to do. I have a wave file. I import it into besweetgui. I choose compliant dvd format wave to mp2. what else is there.

[FredThompson]

BeSweetGui will use toolame and that's what you should do. BeSweet is a very accurate tool for rescaling aspects of audio; sample rate, volume, max volume level, etc.

Why not use DVD2SVCD and chose the SVCD destination but set the resizing and bitrate parameters for D1? Let it automate all the audio conversion for you.

[bb]

Originally posted by troy
what else is there.
Post all your settings, else we can't tell you which one's wrong.

bb

[troy]

Well as I said I used DSPguru template for mp2 for dvd. It sets checkmarks for Use MPA decoder. Toolame as opposed to MP2enc and mp2 is set, Downconverter sample rate is set and hybrid gain is set. Priority class is set to normal and all the other boxes are unchecked. My wave is 48k. I would think that I can just choose this DSpguru mp2 for dvd template and that would take care of everything. I hit wav to mp2 button and then it processes the video. I know that something is wrong with the mp2 because dvd maestro has problems with it too. In dvd lab I get the errors:

Cause bad header or unknown error
This file doesnt seems to be mpeg audio
check if it isn not a wrongly named ac3

[aleksander]

@ bb
Just a small correction.
The forumla that you have used for calculating the video bitrate is slightly wrong. I was doing some DV-DVD lately and it kinda got me confused.

According to you :

available space (in MB) * 1024 / video duration (in sec)
so : 4319 * 1024 / 6250 = 707.62496 (which is not 5660) :)

It should be:

available space (in MB) * 1024 * 8 / video duration (in sec)
so : 4319 * 1024 * 8 / 6250 = 5660.99968

So, the only thing missing is the multiply by 8 thing.
Other than that the guide is great.

take care

aleksander

[bb]

Thanks aleksander. Of course we want to know the bitrate in kilobits per second, not kilobytes.

I updated the guide.

bb

[chriskindom]

I have problem enconding with besweet, using ac3machine and besweetgui
the encode is OK, but when i try to play using powerDVD and WinDVD, there is no sound, (when i compile my dvd with maestro and i try to reproduce the whole DVD there is the same problem, the info say that there is a ac3 sound, but i can´t hear any sound), the file its oka because with another programs., i can hear the AC3.
So for DVD i recommend the vegas AC3 sourround plugin, works great, very simple, and the quality it´s very good.
This is my personal experience, i expect it can be usefull.

SORRY ABOUT MY ENGLISH!!:D





www.killerhog.com.ar
(my film will be in BERLIN in 15 days, in the "b film fest")

[bb]

What were your BeSweet settings?

bb

[chriskindom]

my wav is at 44100hz, stereo, 107 minutes
ac3 options: bitrate 224 channel mode 2.0 SSRC 48000
nothing strange

Christian
www.killerhog.com.ar

[troy]

Is it possible for someone to take some dv footage, demux the audio to wav, and then convert it to mp2 with besweet and see if it is compliant mp2 file.

[Bordo32]

Many thanks - this guide is really great!

@bb, I don’t want to be picky, shouldn’t be a Timecode Setting (in the CCE Encoder setting window) suggested set to 00:00:00:00 instead of keeping at default 01:00:00:00 ?

[bb]

Originally posted by chriskindom
my wav is at 44100hz
Sorry for replying late:
For DVD the sampling rate must be 48000 Hz. Set AC3Machine (or BeSweetGUI) accordingly; upsampling is done through ssrc.

bb

[bb]

Originally posted by Bordo32
@bb, I don’t want to be picky, shouldn’t be a Timecode Setting (in the CCE Encoder setting window) suggested set to 00:00:00:00 instead of keeping at default 01:00:00:00 ?
You're probably right... Does anybody know of the reason why CCE defaults to 01:00:00:00?

bb

[chriskindom]

Originally posted by bb
Sorry for replying late:
For DVD the sampling rate must be 48000 Hz. Set AC3Machine (or BeSweetGUI) accordingly; upsampling is done through ssrc.

bb



yes!, i use the option in AC3machine to upsampling the wav, and i have SSRC. i read carefully all the tutorials. I´m sure theres anything i forgot but i can´t find it.
But when the compression is finish, in PowerDVD and WINDVD there is no sound. I tried many ways.., with vegas surround plugins, i have no problems.
Well, thank you!.

Christian

[NinjaOne]

How do you install Convolution3D I put it in the avs folder but with cce I still get an "Script Error: there is no function named "Convolution3d""

[bb]

Either put it in the "plugins" subdirectory below the AviSynth directory, or use the "LoadPlugin" command in your AVS script.

bb

[NinjaOne]

Thanks for your help, I just finished with the encoding, everything went great.

[matzed]

bb,

Maybe it's a stupid question, but why did you mention in your guide to put "Max bitrate" at 8000 kbps ? I thought that DVD can have a max bitrate at 9800 kbps. With audio at 192 or 224 kbps, having the max bitrate at 9000 would give better quality and leave enough room for titles etc... ?

Matz

[bb]

@matzed:
Because some players have problem with playing DVD+-R media when driven at the limits. So it's recommended to go somewhat lower with the bitrate. If you don't have problems using high bitrates, you can of course go up to 9800 kbps (video and audio and subs altogether). But make sure you don't exceed the max. bitrate.

bb

[matzed]

Thanks, bb. I'll try to see if my DVD player will handle 9.8 Mbps.

Matz

[arkoko]

Sorry my stupid question, but why the following code in script:

Interleave(evn,odd)
Weave()
DoubleWeave.SelectOdd()
Why not only Weave()?

AKo

[FredThompson]

field order is bottom-first with DV. DVD is commonly top-first. This is a way to change the field order but it isn't the best way. It works but, if you examine the stream, you'll see the problems. Simply switching the field order does not work because there is a temporal relationship between the fields.

I don't do this. Most DVD players will accept MPEG2 with bottom field first. Thus, there is no reason to do this operation and it's just one more "change" to the source, reducing quality.

[DDogg]

Hey Fred, how goes the battle in colorspace? :)

I wanted to throw this script up for comment and critique.

I had a situation arise where I needed to get 62 minutes of DV on one 800 cdr and I did not think using straight interlaced at that bitrate was a good idea (even with my beloved ProCoder). This script was the result of snooping around a little. A little of Fred, a little of Stickboy (http://www.avisynth.org/stickboy/), and some scharfis_brain (http://home.arcor.de/scharfis_brain/60ito24p.html) flavored all up with a little dab of Peachsmoother (http://forum.doom9.org/showthread.php?s=&threadid=58674) and stirred with RoBA cooked up a quite decent 24p encode. 62 mins of this on one 800 cd encode in CCE with a Q of 36 and resulting Q average of 5.25 with a peak of 7.25. Granted it was all talking heads stuff, but still I think it worked out pretty well. Even the pans and zooms were decent.

Note: forget this if you have a slow processor, also, you have to apply pulldown with pulldown.exe after encoding.

#Presents TFF
Import("G:\convert60ito24p.avs")
avisource("D:\NTSCDV.avi").DoubleWeave.SelectOdd.ConvertToYUY2( interlaced=true).ReInterpolate411()
jdl_UnfoldFieldsVertical(flip=true)
PeachSmoother(NoiseReduction=50, Stability=30, Spatial=200, NoiseLevel=3.5, BaseLine=2.5)
jdl_FoldFieldsVertical(flip=true)
DGBob(Order=1,mode=1).convert60ito24p(2,0)
#resized to 480x480

/Added - So Fred won't beat me up :)

#Presents BFF (At least I think so)
Import("G:\convert60ito24p.avs")
avisource("D:\NTSCDV.avi").ConvertToYUY2(interlaced=true).ReInterpolate411()
jdl_UnfoldFieldsVertical(flip=true)
PeachSmoother(NoiseReduction=50, Stability=30, Spatial=200, NoiseLevel=3.5, BaseLine=2.5)
jdl_FoldFieldsVertical(flip=true)
DGBob(Order=0,mode=1).convert60ito24p(2,0)
#resized to 480x480

[FredThompson]

The battle continues... ;)

Personally, I don't try to swap field order anymore. You'll either lose the very beginning and end or mess up the temporal aspect. DVDs can be bottom field first. So...why swap, goof up the video, and increase processor time?

Yeah, you've convinced me about ProCoder. Very nice. Very long compiole time but very nice.

KVCD has some stuff you might be interested in. I'd taken their filters and made a field-oriented version with no blurring on the frame before a scene change. That might be around the forums somewhere. Since you're going to progressive, take a look at their stuff. Worth a shot.

[bb]

Originally posted by FredThompson
So...why swap, goof up the video, and increase processor time?
Because you want to encode with CCE?

bb

[FredThompson]

CCE has an advanced setting to set field order

[bb]

Originally posted by FredThompson
CCE has an advanced setting to set field order

:confused:
Which one?

bb

[FredThompson]

I know it's in 2.66+, look at the PDF manual. Maybe Custom's site has a changelog that will show when this became available.

[arkoko]

Originally posted by bb
:confused:
Which one?

bb
In v.2.67 Offset Line=1 = bottom field first.

Ako

[bb]

Yeah, that's the "Upper field first" flag in CCE versions before 2.67. Of course I know that one, but that's not what I'd call an advanced setting to set field order

bb

[FredThompson]

Sigh...it's detailed in the manual in section 3.6, Advanced Settings of the Encoder.

[DDogg]

CCE faq Q11 (http://forum.doom9.org/showthread.php?s=&threadid=53770)

[bb]

Thank you very much, guys. You're telling me stuff that I've know for ages. Thus it's not very helpful, neither the repitition nor the sighing.

From the CCE 2.66 manual:
3.6.3 Specifying field order
If Upper field first is selected, the upper field is displayed on a
TV monitor first. Whether this should be selected depends on the
configuration of the original file to be encoded, but if this is uncertain,
select upper field. If this setting is incorrect, the movement of a
encoded video will be jerky.
This setting is valid only in MPEG-2 encoding mode.

Switching on the upper field checkbox results in the first line of the frame being dropped and the last one being copied. If you prefer this method over the DoubleWeave.SelectOdd() method, it's your choice. I still would not call this an "advanced" method.

bb

[FredThompson]

****, *****. The documentation has the setting in an area called, "advanced." DVD2SVCD has it in the "advanced settings" pop-up. That's all there is to it.

[kempfand]

The best solution is to change field order with Avisynth script, for example
AviSource("V:\original.avi")
DoubleWeave().SelectOdd() and then to CCE-encode with deactivated tff-checkbox.

Checking the tff-checkbox with CCE 2.66+ results in moving the picture one row upwards, in other words: the picture is moved upwards and does no longer correspond to the original. Also, more serious, it doesn't work on every PC.

Regards,
Andreas

P.S. Only applies when no spatio-temporal filters are used, in which case UnfoldFieldsVertical / FoldFieldsVertical should be used

[FredThompson]

Doesn't that give you a half-frame at the beginning and a half-frame at the end? IOW, you get one field in the first frame along with a blank field and the same in the last frame? I seem to recall that happening with the result that it would mess up frame-accurate editing.

Surely, there's a proper way to encode as BFF and just leave it as that, isn't there?

[matzed]

I don't understand everything concerning the last subject (fileds of deinterlaced sources), but for me, when i tested the AVISynth script of the guide, scenes were jerky on my standalone.

I posted here a thread and bb suggested me to drop the last line, DoubleWeave().SelectOdd(), and it worked.

Matz

[bb]

To summarize, how to treat interlaced DV video best regarding the field order depends on the capabilities of your encoder. CCE doesn't support bff encoding, so you have to either use the internal function, which shifts the lines, or use DoubleWeave.SelectOdd(). I prefer the latter. If your encoder supports bff encoding, then do that (but be careful that its bff option doesn't simply shift the lines like CCE does while still encoding tff).

bb

[DDogg]

bb, I finally got off my butt and did a thing on how to create a template in cce 2.67. (http://forum.doom9.org/showthread.php?s=&threadid=64800) Maybe it will be of use to you.

[bb]

DDogg,

excellent. I'll see when I get the time to integrate it into the "DV to DVD" guide. I was thinking of a chapter about the DVD2SVCD automated method for beginners.

bb

[FredThompson]

You might also want to include something about the current release of AVI2SVCD's CVD settings being broken.

[bb]

Originally posted by FredThompson
You might also want to include something about the current release of AVI2SVCD's CVD settings being broken.
Seems like I have to read up on that one...

bb

[FredThompson]

It's been a while but I think the post is in the "beginner's" area for DVD2SVCD. I noticed you could set the output as CVD but that's not what was being generated. IIRC, it was correctly modifying the source resolution but then adding black borders on the sides to yield full D1.

[DDogg]

it should be fixed next release. I am pretty sure it is "in the hopper".

[bb]

*bump*

bb

[RathO]

In the guide, i would say a word on the new Frameserver 2.0, its support is better. Didn't use to work with my MSP7, but now it does!

It works great, is free, and give us the opportunity to encode with the best encoders available. ;)

Here: http://www.debugmode.com/frameserver/

Enjoy!

[BizUser]

First of all, thanks for the guide!

In section 3.1 (CCE), you have '3.1.1 - Create an Avisynth Script'. Is this avisynth script specifically for CCE, or should this be a more general step, regardless of whether you use CCE, TMPG, etc? I was hoping this was the case, because I'd like to take advantage of the presumed benefits of Convolution3D and ReInterpolate411, but I use TMPG, not CCE.

Unfortunately, I see that in section 3.2 (TMPGEnc), you use the avi file as the source, not the avs file, so ... I guess the Avisynth script really is for CCE only. How come?

I tried to use the avisynth script in TMPG, but could not get it to work.

Just in case it should/could work, here's what happened. I first of all realised that I had a very old version of avisynth (have been away from this hobby for almost a year, and used to do DVD backup, but now I need to do DV capture). So I downloaded Avisynth 2.5.

I then downloaded the ReInterpolate411 plugin, but got an error that it required YUY2 input (confirmed in the reinterpolate411 information that it requires YUY2 format for Avisynth 2.5 at http://mywebpages.comcast.net/trbarry/downloads.htm). I set the pixel_type to YUY2 in the AVISource string, and solved that problem.

But then I downloaded the convolution3D filter, and when I used it in my script, got an error that it was not compatible with Avisynth 2.5. Found an updated version (beta ...) of convolution3d (http://www.hellninjacommando.com/con3d/beta/) for avisynth 2.5, but then got an error that convolution3d requires YV12 - directly contradicting the need for YUY2 above (or am I missing something here?). Am I required to do a (second?) color space conversion (from YUY2 to YV12 before invoking convolution3D?

My source AVI was created by ScenalyzerLive, using the type 2 canopus codec, in case that matters. Thanks!

<update> - I found a version of convolution3d that works with YUY2, and I'm testing with it now. The curious thing is, it's not on the author's own site... I had to get it from another site. Anyway, if you run into this issue, you can go to this site: http://www.avisynth.org/warpenterprises/
and download the yuy2 version - it's the one that does NOT have a colorspace designation in the name (the other one has 'yv12' in the name). I figured this out from this guide http://www.videohelp.com/forum/viewtopic.php?t=225951 over at videohelp.com.

[bb]

Originally posted by BizUser
[...]In section 3.1 (CCE), you have '3.1.1 - Create an Avisynth Script'. Is this avisynth script specifically for CCE, or should this be a more general step, regardless of whether you use CCE, TMPG, etc? I was hoping this was the case, because I'd like to take advantage of the presumed benefits of Convolution3D and ReInterpolate411, but I use TMPG, not CCE.
Yes, you can use AviSynth in conjunction with other encoders. You may want to modify the script under certain circumstances, e.g. not convert to upper field first video, and encode bff instead (provided that the encoder supports that).

Originally posted by BizUser
[...]Unfortunately, I see that in section 3.2 (TMPGEnc), you use the avi file as the source, not the avs file, so ... I guess the Avisynth script really is for CCE only. How come?
TMPGEnc can do without AviSynth, although it's still a good idea to use it because of the many great filters etc. TMPGEnc has internal filters you may use, and I guess for beginners it's easier to use TMPGEnc alone.

Originally posted by BizUser
I tried to use the avisynth script in TMPG, but could not get it to work.
Check for the AVI reader priority. I assume that you'll find a solution in the forum by using the search button, too ;)

Originally posted by BizUser
[...]I then downloaded the ReInterpolate411 plugin, but got an error that it required YUY2 input [...] Am I required to do a (second?) color space conversion (from YUY2 to YV12 before invoking convolution3D?[...]
Check out http://www.avisynth.org/warpenterprises/
Use convolution3d_25_dll_20030422.

bb

[BizUser]

Originally posted by bb
Yes, you can use AviSynth in conjunction with other encoders. You may want to modify the script under certain circumstances, e.g. not convert to upper field first video, and encode bff instead (provided that the encoder supports that).
bb
After downloading the correct color-space version of convolution3D, I was able to get everything to work (I found it and updated my post, presumably while you were writing your reply!).

However, the output was clearly messed up - interlaced wrong (jerky).

Now I see your reply, and you say '...not convert to upper field first video, and encode bff instead'. I figured that out myself, and changed TMPGEnc from BFF (the setting I originally used since it wast DV source) to TFF and solved the problem. But I'd rather remove un-necessary stuff from the avisynth script. To this end, I re-read your guide and all the posts in this thread. I have just one question ... the following three lines apparently switch field order to TFF:


Interleave(evn,odd)
Weave()
DoubleWeave.SelectOdd()

One post implied that this could be replaced by just 'Weave()', but another suggests that you would simply remove the last line ('DoubleWeave.SelectOdd()'). So ... for TMPGEnc, with a BFF DV source AVI, what exactly should I use to produce BFF input to TMPGEnc? I'm guessing the Interleave(evn,odd) has to be there to re-constitute the interlaced material, but would like to double-check!

[Peter1234]

BizUser,

DoubleWeave.SelectOdd() #### this reverses field order

Interleave(even,odd).Weave() #### this re-interleaves fields

[bb]

Yes, to keep the video BFF, just remove the line DoubleWeave.SelectOdd()

bb

[Video Dude]

I read the debate on pages 1 and 2 about the colorspaces (0-255 and 16-235). But that discussion was based on RGB conversions.

What if your DV codec decodes in YUY2 0-255?

Should you add Limiter() at the end of the AviSynth script to clamp it to 16-235 if your goal is DVD for display on a TV?

The luminace selection buttons in CCE have no effect since it is YUY2 input and not RGB (according to both Doom9 and RB's CCE guides). So should CCE recieve YUY2 0-255 or YUY2 16-235 for DVD encoding?

The AviSynth Docs suggest YUY2 16-235 for TV display, but I just want to confirm this as Limiter seems to make the video brighter.

[Wilbert]

IF it gives YUY2 0-255 (which you can check by using the histogram, and looking up some very bright/dark spots), then you should scale it to 16-235 (and not clamp to it) using

ColorYUV(levels="PC->TV")

[bsr]

If I'm not going to do any filtering, but I want to resize to 352x480, would the prefered way be to just use TMPGEnc, crop to 704 and encode to 352.
Or should I use AVISYNTH for crop to 704x480, Lanzcos Resize to 352x480, and then use TMPGEnc to just encode?
Or does it not really make a difference, just looking for others preferences.

[bb]

bsr,

I don't believe that you'd see much of a difference. Lanczos resize is more accurate than Bicubic, but in my opinion it's not worth the extra rendering time for video (for stills it's another story).

I'd use AviSynth if I need special filters, and TMPGEnc alone if not.

bb

[Kika]

ColorYUV(levels="PC->TV")

If TMPGEnc is used for Encoding, there's absolutly no need to do this. TMPGEnc is able to handle this correctly.

For Source with a Lumarange of 0-255, got to Setting -> Quantize matrix and set Output YUV data as basic YCbCr not CCIR to the OFF-State. If the Range is 16-235, set it to ON.
This will give you much better results than working with ColorYUV.

[Wilbert]

If TMPGEnc is used for Encoding, there's absolutly no need to do this. TMPGEnc is able to handle this correctly.
Yeah, yeah. I still remember the discussion :) He was talking about CCE.

[Kika]

Yeah, but bb talked about TMPGEnc ;)

BTW: Is the actual Version of CCE able to handle this Luma-Range-Stuff correctly? Or does it still need ColorYUV?

[bb]

Are we still on topic? I'd appreciate some more comments on the guide itself.

bb

[Kika]

OK, you wanted it... ;)

I suggest you to insert a Note about the Field Order Stuff. Like others here said, your Script reverses the Field order. But in the TMPGEnc-Selection you are using BFF, which is the wrong setting if your Script is used.
You have not to change the Script or the Guide at all, just insert a couple of notes to prevent misunderstandings.

Kika

[Wilbert]

BTW: Is the actual Version of CCE able to handle this Luma-Range-Stuff correctly? Or does it still need ColorYUV?
You mean if you feed it with YUV [0,255]? I don't know cce very well, but I suspect it won't scale it. It shouldn't be difficult to test this.

[ZeB]

3.2 TMPGEnc

It is really easy to use TMPGEnc for DV to DVD conversions. Start the encoder, cancel the wizard (if applicable), and open your DV type-1 or type-2 file via the "Browse" button next to the "Video source" field.

If I do this, I just get a black screen in TMPGEnc. It works if I make an avisynth script (DirectShowSource("DV-video.avi", 25)). Even CCE can open the DV-file.

What am I doing wrong?

[Kika]

Raise the Priority of the DirectShow Multimedia File reader VFAPI PlugIn in TMPGEnc (Option -> Environmental setting -> VFAPI Plug-in)

[ZeB]

Thanks Kika! works like a charm now.