The "original" guide (v1.0) was written in German language and is hosted
at doom9.de, the German mirror of doom9.org. It was written by BaronVlad
(who sadly stopped being an active moderator/member of doom9, and continued
to work on this guide till v2.0) and translated by Steve.
For the second (v2.0) version some people wrote new parts for this
guide. Wilbert wrote the AviSynth and GKnot part and Steve wrote the logo
removal part. We were able to add some words about NTSC capturing and the
neverending story "What is Full PAL/NTSC ?". This was a hard nut to crack
and we wouldn't have been able to crack it without Karl, Wilbert, theWEF
and other helping hands.
You are now reading the third (v3.0) version of the guide. We had many
good suggestions, and a lot is added. The PAL/NTSC resizing part is improved,
making a clear difference between 'ITU compliant capturing' and 'capturing
with a horizontal scaler'. The VirtualVCR capturing (based on WDM) part
is added by zisoft and translated by Steve. The BT8x8 Tweaker part is added
by ppera2. Regarding AviSynth: the logo removal part is extended, removing/correcting
of rainbows and chromashifts is explained. We tried to make this guide
a bit more technical than the previous versions. Although this will please
some people, it will also be a bit harder to follow for people who just
began with capturing. We hope that we have found a good mix.
We had to add these words here to let you know, that this guide is
not the work of one person! Thanks for your help. More information can
be found in the Appendix.
The orginal author BaronVlad comments "This guide is merely a guideline and definitely not the only way to achieve the desired results. Nor is it by any means complete, nor does it contain all information existing on this topic. It is only one approach, and provides some tips people have found useful for capturing to their liking. When I started putting my time into this topic (which wasn't very long ago) I wasn't able to find a complete guide, probably because I didn't know where to look. I want to save everybody else having to "reinvent the wheel"."
In this guide you will be able to learn how you can capture analog video material to your computer. The source will be a PAL or a NTSC system and you should end up with a DivX .avi file muxed with a cbr-mp3 audio, in your preferred filesize. For the more advanced users, who prefer vbr-mp3 or ogg-vorbis, a small excursion into high quality sound is included.
Additionally, I discuss how to burn your finished videos to Video CD or SuperVideo CD. I will just treat resolutions differing from the standard and frameserving with VirtualDub in another application. The coding as well as the burning itself you can easily get from the numerous other guides on this webpage.
The programs used for capturing is VirtualDub (VfW based) and VirtualVCR (WDM based). It allows you to achieve better results than with the mpeg2 capture software provided with your capture card. We separated the tasks of capturing and storing video. This means that you first save the material (including commercials and all the other junk you'll later want to get rid of) in almost lossless compression to your hard drive, and afterwards you deal with the task of processing and converting it to its final format. We are going to describe three possibilities for capturing. The choice is solely up to you, since "best" is what works best for you. Nevertheless, some support for your decision making is provided.
Postprocessing the captured video will be explained in three different ways: VirtualDub and GKnot as the "easy" way and AviSynth postprocessing. In the beginning AviSynth might scare you, but you should take a look into the AviSynth part ASAP to learn this simple scripting language, because it will speed up the encoding process and will give you very good results
You have three important decisions to make:
I have to mention though that the MJPEG codec produces a picture that is a little less sharp, which may not fit everybody's needs. In addition, the codec is not free; you have to register by paying a fee to avoid watermarks being put onto the video (we are going to handle them later on though).
When handling such large files, both the processor and your hard disk will incur a significant work load. Therefore when choosing the codec, you should also consider your hardware. You'll need a fast processor and a fast hard disk [translator: in my opinion, for (almost) lossless video, a one hour plus video requires at least 40 gig of disk space]. But since the file size also depends very much on the resolution you're choosing, you should also consider your final choice:
If you know about dvd resizing and you think that resizing analog captures works the same way, then you should read this section very carefully. Because it depends on your capture card whether this is true. You will encounter one of the following two situations:
Make two captures, one at 704x576 and one at 720x576 (or 704x480 and 720x480 for NTSC). If the first capture is a scaling of the second, it implies your capture card/chip performs internal scaling. But if the second capture has approx.16 pixels more overscan (that is vertical black borders; 9 pixels for NTSC), it implies your capture card/chip is ITU compliant. (Capturing at the resolution 720x576 or 720x480 is not sufficient to see this, since the TV transmission itself can also include vertical black borders.)
References:
BT 8x8
Data Sheet: Have a look at the 100119a.pdf document.
Conexant
cx2388x Data Sheet
Philips SAA7108
/ 7113
Data Sheets
Since TV transmission is analogue, there are no pixels. But one needs pixels for digital storage. Therefore one needs a guideline, which prescribes how the analogue signals have to be converted into computer pixels. This guideline (as recommendation) is given in the Recommendation ITU-R BT.601-5. Some important issues:
As already mentioned, TV transmission is analogue. The process of digitizing it by the capture card, is called sampling or performing an analogue to digital conversion. Mathematically it just means that a waveform is discretized in a certain number of parts, and these parts are called samples. Looking at "page 13 of the BT 8x8 Data Sheet" it is mentioned that the number of samples is always the same (NTSC: 910, and PAL: 1135 (4)), independently of the chosen capture resolution. After sampling, the clip is resized to the resolution which you used when capturing (with some cards, overscan is included).
The reason of this high number of samples, is the fulfilment of Nyquist sampling. Have a look at this thread for more information.
In other words: if you capture in a higher resolution, the capture will be more precise, but not due to oversampling. It is just because you are downsizing to a less smaller resolution. So, in general, capture at as high a resolution as possible. Filter before resizing so you have as much data as possible for the filtering. Resize after that with a method which you know is high quality to reduce distortion as much as possible.
384x288 (1/4 PAL) = Low Quality
+ low hardware and software requirements
+ requires the least hard disk space
+ Huffyuv is also usable when choosing smaller hard disks
+ deinterlacing not necessary
+ filtering IMO not necessary (for the reason I explained above), but
if you want to filter the video the same filters will work faster because
of less informations to be processed
+ takes less time to convert the captured file to the desired format
- less information is being captured
- modest video quality
384x576 with vertical resize ("1/2 PAL", not scientifically correct,
but clear I hope) = Average Quality
Basically only half of the PAL width (768/2 = 384), but the complete
height is captured. But also the height is divided by two during the capture
process so the resulting resolution will be at 384x288.
+ smaller requirements for the hard disk (compared to Full PAL, though
higher requirements for the CPU)
+ requires less hard disk space
+ Huffyuv also usable for smaller disks
+ still no deinterlacing necessary
+ almost no use of filters
+ faster processing of the files later on
- more information saved than in 1/4 PAL but not nearly as much as
in full PAL
- average video quality
704/720/768x576 (Full PAL) = High Quality
What exactly is Full PAL ? We are talking about analogue capture, in
other words take an analogue source and digitalize it. Information in analogue
sources is not stored in bits and pixels and so your capture card has to
be told to create a pixel resolution so your computer can further process
it. But we normally have a DAR (Display Aspect Ratio) of 4/3, on the TV
screen AND on the computer monitor. As a standard we always have 576 active
scan lines. This should result in 576 * 4/3 = 768.
In case your capture card performs horizontal scaling the resizing process
is very simple: don't crop, and resize to a 4:3 format (like 640x480).
If you want to use FitCD and GKnot to determine your resize settings, it
becomes more difficult. If you captured at 720x576, you have to use a generic
PAR (Pixel Aspect Ratio). They are listed in the following table:
| capturing: horizontal scaling | generic PAR: |
| 704x576 | 12/11 (accepting a small error, use PAR = 1/1) |
| 720x576 | 48/45 |
| 768x576 | 1/1 |
In case your capture card is ITU compliant the resizing process becomes
more difficult. Der Karl's calculations (Der
Karl's Apect Ratio for Dummies) give us a result of approx. 702 (active,
horizontal) pixels for the right resolution, which gives us a PAR of 128/117
(note that 768/702 = 128/117). It's outside the scope of this guide to
explain this in detail here. It follows that if you captured at 704x576
or 768x576 you can resize directly to a 4:3 format (like 640x480).
If you captured at 720x576, it implies that you have to crop the overscan
away (ending with 704x576) and resize to a 4:3 format (like 640x480).
If you want to use FitCD and GKnot to determine your resize settings, the
PAR's are given in the following table:
| capturing: ITU compliant | cropping or adding black bars | PAR: |
| 704x576 with 2 pixels overscan | 702x576 (accepting a small error, leave it at 704x576) | 128/117 |
| 720x576 with 18 black pixels added horizontally | 702x576 (accepting a small error, crop to 704x576 instead) | 128/117 |
| 768x576 (1) | - | 1/1 |
There are two ways to get rid of the horizontal black borders (in case your capture contains them as is often the case). The easiest is to crop overscan and resize first, and then crop away the horizontal black borders. Note that your final clip will not be 4:3 anymore. The hardest way is to crop away the horizontal black borders first, and then resize to a correct format. If you want to do this correct, you have to use the appropriate PAR's. Since the first method is much easier and you will not have much black borders anyway, I suggest to use the first method (that will be done in this guide).
The advantages/disadvantages of capturing at Full PAL:
+ all possible information is captured
+ best quality possible
- very high requirements for hardware and software
- huge amounts of hard disk space needed
- Huffyuv can't be used with smaller hard disks
- deinterlacing is necessary
- generally requires a lot of filtering
- requires large processing power and time
Other Formats (VCD: 352x288 or SVCD: 480x576)
If you want to create a (S)VCD the resolution MUST be 352x288 (VCD)
or 480x576 (SVCD). Standalones will not support any other resolutions because
these are fixed standards. To achieve these formats there are two possibilities:
1) The video can be directly recorded in the required resolution. This
would result in lower quality, because you do not capture all the information
available to help the filters (also resize filters) to do their job. Therefore,
it is always better to capture at Full PAL and do the resizing later, if
your hard- and software admits it. However, these different resolutions
are possible. If you choose to directly capture at 352x288, please follow
the 1/4 PAL settings. If you instead choose SVCD or plan on resizing later,
you should use the Full PAL sections of my guide. If there is a need for
a departure from any standard I will tell you to do so at the right moment.
Remember, the choice of codec (quality vs. processing time) can only be made by you. Please consider your hardware well. It will NOT be possible to record three hours of Huffyuv with your PII 400 and 10 gig hard drive.
You can postpone the final decision for a few more minutes, because the next step will be just about optimizing your system, getting any programs you might need later on, and adjusting the basic settings.
320x240 (1/4 NTSC) = Low Quality
+ low hardware and software requirements
+ requires the least hard disk space
+ Huffyuv is also usable when choosing smaller hard disks
+ deinterlacing not necessary
+ filtering IMO not necessary (for the reason I explained above), but
if you want to filter the video the same filters will work faster because
of less informations to be processed
+ takes less time to convert the captured file to the desired format
- less information is being captured
- modest video quality
320x480 with vertical resize ("1/2 NTSC", not scientifically correct,
but clear I hope) = Average Quality
Basically only half of the NTSC width (640/2 = 320), but the complete
height is captured. But also the height is divided by two during the capture
process so the resulting resolution will be at 320x240.
+ smaller requirements for the hard disk (compared to Full NTSC, though
higher requirements for the CPU)
+ requires less hard disk space
+ Huffyuv also usable for smaller disks
+ still no deinterlacing necessary
+ almost no use of filters
+ faster processing of the files later on
- more information saved than in 1/4 NTSC but not nearly as much as
in full NTSC
- average video quality
704/720/640x480 (Full NTSC) = High Quality
What exactly is Full NTSC ? We are talking about analogue capture,
in other words take an analogue source and digitalize it. Information in
analogue sources is not stored in bits and pixels and so your capture card
has to be told to create a pixel resolution so your computer can further
process it. But we normally have a DAR (Display Aspect Ratio) of 4/3, on
the TV screen AND on the computer monitor. As a standard we always have
486 active scan lines. This should result in 486 * 4/3 = 648. However,
the main difficulty is that 6 scan lines are cropped of during capturing
(the data sheets don't mention this, but since the quality would degrade
much when scaling instead of cropping those 6 lines, I assume that the
6 lines are cropped), and we have to compensate for this.
In case your capture card performs horizontal scaling the resizing process
is very simple: don't crop, add black borders of six pixels vertically
(to end up with xxxx486) and resize to a 4:3 format (like 640x480).
If you want to use FitCD and GKnot to determine your resize settings, it
becomes more difficult. You have to use generic PAR's (at least if you
captured at 720x480), which are listed in the following table:
| capturing: horizontal scaling (2) | generic PAR: |
| 704x480 | 81/88 |
| 720x480 | 9/10 |
| 640x480 | 81/80 (accepting a small error, use PAR = 1/1) |
In case your capture card is ITU compliant the resizing process becomes
more difficult. The ITU-R
BT.601-5 standard
gives us a result of approx. 711 (active, horizontal) pixels for the right
resolution, which gives us a PAR of 72/79 (note that 648/711 = 72/79 and
648/486 = 640/480). It's outside the scope of this guide to explain this
in detail here. It follows that if you captured at 704x480 or 640x480 you
can resize directly to a 4:3 format (like 640x480). While capturing
at 720x480, implies cropping the overscan away (ending with 704x480)
and resize to a 4:3 format (like 640x480). If you want to use FitCD
and GKnot to determine your resize settings, the PAR's are given in the
following table:
| capturing: ITU compliant (assuming that the 486-480=6 vertical lines are used) (3) | cropping or adding black bars | PAR: |
| 704x480 with 2 pixels overscan | 702x480 (accepting a small error, leave it at 704x480) | 72/79 |
| 720x480 with 18 pixels overscan | 702x480 (note: YUV requires even widht/height, accepting a small error, crop to 704x480 instead) | 72/79 |
| 640x480 | - | 1/1 |
There are two ways to get rid of the horizontal black borders (in case your capture contains them as is often the case). The easiest is to crop overscan and resize first, and then crop away the horizontal black borders. Note that your final clip will not be 4:3 anymore. The hardest way is to crop away the horizontal black borders first, and then resize to a correct format. If you want to do this correct, you have to use the appropriate PAR's. Since the first method is much easier and you will not have much black borders anyway, I suggest to use the first method (that will be done in this guide).
The advantages/disadvantages of capturing at Full NTSC:
+ all possible information is captured
+ best quality possible
- very high requirements for hardware and software
- huge amounts of hard disk space needed
- Huffyuv can't be used with smaller hard disks
- deinterlacing is necessary
- generally requires a lot of filtering
- requires large processing power and time
Other Formats (VCD: 352x240 or SVCD: 480x480)
If you want to create a (S)VCD the resolution MUST be 352x240 (VCD)
or 480x480 (SVCD). Standalones will not support any other resolutions because
these are fixed standards. To achieve these formats there are two possibilities:
1) The video can be directly recorded in the required resolution. This
would result in lower quality, because you do not capture all the information
available to help the filters (also resize filters) to do their job. Therefore,
it is always better to capture at Full NTSC and do the resizing later,
if your hard- and software admits it. However, these different resolutions
are possible. If you choose to directly capture at 352x240, please follow
the 1/4 NTSC settings. If you instead choose SVCD or plan on resizing later,
you should use the Full NTSC sections of my guide. If there is a need for
a departure from any standard I will tell you to do so at the right moment.
Remember, the choice of codec (quality vs. processing time) can only be made by you. Please consider your hardware well. It will NOT be possible to record three hours of Huffyuv with your PII 400 and 10 gig hard drive.
You can postpone the final decision for a few more minutes, because the next step will be just about optimizing your system, getting any programs you might need later on, and adjusting the basic settings.
Footnotes:
(1) It is assumed here that capping at 768x576 is not ITU compliant,
in other words the full 4:3 image is captured. I haven't seen any evidence
for this. So if you can capture at 768x576 with an ITU compliant capture
card, please drop me a mail!
(2) As explained in the text, it is assumed that the six scan lines
are cropped of during capturing. It is also assumed that there is no horizontal
compensation for this (thus the actual image is.1.35:1). I have no evidence
for this.
(3) As explained in the text, it is assumed that the six scan lines
are cropped of during capturing. Here, it is also assumed that there IS
a horizontal compensation for this (thus the actual image is.4:3). I have
no evidence for this. So if you have an ITU compliant capture card (NTSC),
drop me a mail!
(4) This appears only to be the case for BT 8x8 and cx2388x chipsets.
Next step: optimizing your system and software needed: <NEXT>
Back to the Start: <HOME>
English version last edited on: 09/17/2003 | First release: n/a | Authors: Wilbert & BaronVlad | Translator:SteVe(killingspree) | Content by Doom9.org