lancer
28th December 2002, 19:32
okay, after a weeks playing around getting to grips with avisynth scripts, virtual dub and tmpgenc I've arrived at my solution to capture a 2 hour film on windows98. i'd load win 2000 but my capture port doesn't work under 2000, more's the pity.
my rig.
Intell PIII 733
512Mb SDRAM
Asus Geforce 1 Deluxe 6800 32mb ddr with composite capture port.
Maxtor UDMA 133 80GB hard drive on a udma 66 controller partitioned into 2 40GB drives
software
Avisynth 2.07
virtual dub latest
TMPGenc 2.56
using some of donald grafts virtualdub scripts and his decomb filter to IVTC the NTSC original into a progressive 24fps source for the dvd.
the way I see it I have two ways to capture and encode but I'm concerned about recompression and loss of quality. see my methods below.
if I capture, and then frameserve these capture files to tmpgenc I am only compressing once, the time the initial lossless capture is converted to mpeg.
some people have said on the forum to run filters and make interim files, but if I'm going from the original capture, making an interim file and them compressing that to mpeg, wouldn't I lose quality? because I'm recompressing a file that's already been recompressed.
second, frameserving while running filters. should I do so? because surely that's going to make the process longer. of course at the same time if make an interim file first, then I have the time that takes as well as the time it takes to make the mpeg.
finally which is faster? using the script to resize and crop, or virtualdub? I ask, because when frameserving my tests so far show that running the bicubic resize and crop in the avisynth script is about 50% slower than letting virtualdub do it.
could someone experienced cast an opinion on my proposed methods and suggest anything I may have missed, is there any way to speed it up either in script or by omitting some setting in either tmpgenc or virtualdub etc. at the moment, the frameserve of 1 hours footage to make a 2gb 1 hour mpeg file is going to take about 30 hours because of the 2 pass vbr selection.
method 2
capture using the Asus Live Video vid cap program at 704 x 480. I have to because virtual dub doesn't see my capture port (I'm pissed at this because it used and then one day just stopped)
using Huffyuv set to convert to yuy2 that gets me my 2gb avi's. the program spans so that is okay but I drop about 1 frame in 70-80 frames this way.
then run virtual dub with the following script
LoadPlugin("decomblegacy.dll")
a = OpenDMLSource("side1 capture.avi").BicubicResize(720,640)
b = OpenDMLSource("side 2 capture.avi").BicubicResize(720,640)
c = OpenDMLSource("side 3 capture.avi").BicubicResize(720,640)
d = OpenDMLSource("side 4 capture.avi").BicubicResize(720,640)
e = OpenDMLSource("side 5 capture.avi").BicubicResize(720,640)
f = OpenDMLSource("side 6 capture.avi").BicubicResize(720,640)
AlignedSplice(a, b, c ,d ,e, f)
ConvertToYUY2()
Telecide(guide=1, post=true)
Decimate(cycle=10)
Crop(0, 88, -0, -72)
that gets me a pseudo anamorphic 720 x 480 file. I then use donald grafts smart smoother filter at diameter 5, threshold 25.
I frameserve while running the filters to tmpgenc as a 4:3 525 line NTSC progressive source and set to 16:9 output with a 3:2 pulldown flag set. I use 2 pass vbr with average at 5000 to get my final mpeg file for use on spruce for dvd authoring.
method 2
capture using the Asus Live Video vid cap program at 704 x 480. I use Asus own compression software at quality 9 which gets a very nice looking file and gives me 4 1/2 hour avi's of under 2GB each. this loses no frames at all unlike the huffyuv capture method.
(I've found that if I frameserve direct to tmpgenc with this avi, the sound sync gets progressively worse. eg. 5 minute clip. at the beginning it's in sync, by the end its out by 1 second. over 20 minutes of clip it's out by 5 seconds at the end. so using virtual dubs sound correction won't work.)
that means I need to make an interim file using the filters and segmented avi's via virtual dub. so I then run virtual dub with the following script
LoadPlugin("decomblegacy.dll")
a = OpenDMLSource("side1 capture.avi").BicubicResize(720,640)
b = OpenDMLSource("side 2 capture.avi").BicubicResize(720,640)
c = OpenDMLSource("side3 capture.avi").BicubicResize(720,640)
d = OpenDMLSource("side 4 capture.avi").BicubicResize(720,640)
AlignedSplice(a, b, c, d)
ConvertToYUY2()
Telecide(guide=1, post=true)
Decimate(cycle=10)
Crop(0, 88, -0, -72)
that gets me a pseudo anamorphic 720 x 480 file. I then use donald grafts smart smoother filter at diameter 5, threshold 25.
I save this lot as a segmented avi using Huffyuv. once that's done I then run the following script
a = OpenDMLSource("side1 capture.avi").BicubicResize(720,640)
b = OpenDMLSource("side 2 capture.avi").BicubicResize(720,640)
AlignedSplice(a, b)
for hour one and
a = OpenDMLSource("side3 capture.avi").BicubicResize(720,640)
b = OpenDMLSource("side 4 capture.avi").BicubicResize(720,640)
AlignedSplice(a, b,)
for hour two.
I frameserve these to tmpgenc as a 4:3 525 line NTSC progressive source and set to 16:9 output with a 3:2 pulldown flag set. I use 2 pass vbr with average at 5000 to get my final mpeg file for use on spruce for dvd authoring.
my rig.
Intell PIII 733
512Mb SDRAM
Asus Geforce 1 Deluxe 6800 32mb ddr with composite capture port.
Maxtor UDMA 133 80GB hard drive on a udma 66 controller partitioned into 2 40GB drives
software
Avisynth 2.07
virtual dub latest
TMPGenc 2.56
using some of donald grafts virtualdub scripts and his decomb filter to IVTC the NTSC original into a progressive 24fps source for the dvd.
the way I see it I have two ways to capture and encode but I'm concerned about recompression and loss of quality. see my methods below.
if I capture, and then frameserve these capture files to tmpgenc I am only compressing once, the time the initial lossless capture is converted to mpeg.
some people have said on the forum to run filters and make interim files, but if I'm going from the original capture, making an interim file and them compressing that to mpeg, wouldn't I lose quality? because I'm recompressing a file that's already been recompressed.
second, frameserving while running filters. should I do so? because surely that's going to make the process longer. of course at the same time if make an interim file first, then I have the time that takes as well as the time it takes to make the mpeg.
finally which is faster? using the script to resize and crop, or virtualdub? I ask, because when frameserving my tests so far show that running the bicubic resize and crop in the avisynth script is about 50% slower than letting virtualdub do it.
could someone experienced cast an opinion on my proposed methods and suggest anything I may have missed, is there any way to speed it up either in script or by omitting some setting in either tmpgenc or virtualdub etc. at the moment, the frameserve of 1 hours footage to make a 2gb 1 hour mpeg file is going to take about 30 hours because of the 2 pass vbr selection.
method 2
capture using the Asus Live Video vid cap program at 704 x 480. I have to because virtual dub doesn't see my capture port (I'm pissed at this because it used and then one day just stopped)
using Huffyuv set to convert to yuy2 that gets me my 2gb avi's. the program spans so that is okay but I drop about 1 frame in 70-80 frames this way.
then run virtual dub with the following script
LoadPlugin("decomblegacy.dll")
a = OpenDMLSource("side1 capture.avi").BicubicResize(720,640)
b = OpenDMLSource("side 2 capture.avi").BicubicResize(720,640)
c = OpenDMLSource("side 3 capture.avi").BicubicResize(720,640)
d = OpenDMLSource("side 4 capture.avi").BicubicResize(720,640)
e = OpenDMLSource("side 5 capture.avi").BicubicResize(720,640)
f = OpenDMLSource("side 6 capture.avi").BicubicResize(720,640)
AlignedSplice(a, b, c ,d ,e, f)
ConvertToYUY2()
Telecide(guide=1, post=true)
Decimate(cycle=10)
Crop(0, 88, -0, -72)
that gets me a pseudo anamorphic 720 x 480 file. I then use donald grafts smart smoother filter at diameter 5, threshold 25.
I frameserve while running the filters to tmpgenc as a 4:3 525 line NTSC progressive source and set to 16:9 output with a 3:2 pulldown flag set. I use 2 pass vbr with average at 5000 to get my final mpeg file for use on spruce for dvd authoring.
method 2
capture using the Asus Live Video vid cap program at 704 x 480. I use Asus own compression software at quality 9 which gets a very nice looking file and gives me 4 1/2 hour avi's of under 2GB each. this loses no frames at all unlike the huffyuv capture method.
(I've found that if I frameserve direct to tmpgenc with this avi, the sound sync gets progressively worse. eg. 5 minute clip. at the beginning it's in sync, by the end its out by 1 second. over 20 minutes of clip it's out by 5 seconds at the end. so using virtual dubs sound correction won't work.)
that means I need to make an interim file using the filters and segmented avi's via virtual dub. so I then run virtual dub with the following script
LoadPlugin("decomblegacy.dll")
a = OpenDMLSource("side1 capture.avi").BicubicResize(720,640)
b = OpenDMLSource("side 2 capture.avi").BicubicResize(720,640)
c = OpenDMLSource("side3 capture.avi").BicubicResize(720,640)
d = OpenDMLSource("side 4 capture.avi").BicubicResize(720,640)
AlignedSplice(a, b, c, d)
ConvertToYUY2()
Telecide(guide=1, post=true)
Decimate(cycle=10)
Crop(0, 88, -0, -72)
that gets me a pseudo anamorphic 720 x 480 file. I then use donald grafts smart smoother filter at diameter 5, threshold 25.
I save this lot as a segmented avi using Huffyuv. once that's done I then run the following script
a = OpenDMLSource("side1 capture.avi").BicubicResize(720,640)
b = OpenDMLSource("side 2 capture.avi").BicubicResize(720,640)
AlignedSplice(a, b)
for hour one and
a = OpenDMLSource("side3 capture.avi").BicubicResize(720,640)
b = OpenDMLSource("side 4 capture.avi").BicubicResize(720,640)
AlignedSplice(a, b,)
for hour two.
I frameserve these to tmpgenc as a 4:3 525 line NTSC progressive source and set to 16:9 output with a 3:2 pulldown flag set. I use 2 pass vbr with average at 5000 to get my final mpeg file for use on spruce for dvd authoring.