I often run complicated or very slooowwww processes as two separate scripts. Let's say the first half of the routines will deinterlace or ivtc and correct levels, but you want to tweak a few repairs after doing that. Why run the first part over and over when what you're looking for is a proper filter or tweak for something near the end of the processing? If you want, later you can always restore the two parts to one humongous memory hog.
Try StreamFab Downloader and download from Netflix, Amazon, Youtube! Or Try DVDFab and copy Blu-rays! or rip iTunes movies!
+ Reply to Thread
Results 211 to 231 of 231
Thread
-
Last edited by sanlyn; 19th Mar 2014 at 03:19.
-
Some filters that often cause crashes when used in combination:
QTGMC()
MCTemporalDenoise()
RemoveSpotsMC()
ReplaceFramesMC() -
Maybe this post about JVC decks will be helpful: https://forum.videohelp.com/threads/362599-Please-review-my-home-movie-VHS-capture-work...=1#post2305120
-
I believe the script dianedebuda referred to isn't the script I posted yesterday but is from an earlier post. In that case, three of the four mentioned plugins were in use (I don't think MCTD was included). Run something like that 4 or 5 times and start time-skipping back and forth for a while, and you'll get the same or similar error popping up sooner or later. That kind of stress test doesn't surprise me. As I said, I figure I've pushed things too much and will reboot to start fresh.
Stab() is another one that can be added to that combo list.
I had the same experience with my old JVC 7600's.Last edited by sanlyn; 19th Mar 2014 at 03:20.
-
I believe the script dianedebuda referred to isn't the script I posted yesterday but is from an earlier post.
Is there an avisynth method to save video? That is, an output form like AviSource. Looked but didn't see anything. Know it's designed to be a frame server to stuff like vdub which can save, but wondered if there's a "direct" method.
Maybe this post about JVC decks will be helpful
Edit: Guess I should again say I'm running avisynth in avsPmod right now, not as input to vdub. -
No you must use a separate encoder. You can use a batch file and a CLI based encoder to do the encoding though. That's what I usually do (after checking the script with VirtualDub). You can drag/drop the AVS script into the batch file. Or, if you put that batch file in your Send To folder you can right click on the AVS file and Send To -> the batch file.
Here's an example batch file:
Code:start /b /low x264.exe --preset=slow --crf=18 --sar=1:1 --keyint=48 --output %1.mkv %1 pause
-
I use Notepad and VirtualDub.
In theory, it's possible to create a filter that saves the video to a file. I don't know of any that does though. -
Notepad for scripts, VirtualDub to view, edit, filter, save, etc. "Mod' just gets in the way.
Last edited by sanlyn; 19th Mar 2014 at 03:20.
-
Still having troubles with sanlyn's vhs guitar script. Have split into 2 pieces: 1) all processing before vdub filters and 2) the rest. Seems that Interleave is the problem? With AvsPMod, get error popup window; with vdub, seems to load ok & the direct steam avi output file seems ok, but then get vdub crash msg. Don't know if contents of file are good - that is, that Interleave worked fully. It plays.
This causes error:
Code:ConvertToRGB32(matrix="Rec601",interlaced=false) #-----Correct tint in in Even and Odd frames ("e1" and "o1") to make them more alike. AssumeTFF() a=last e1=a.SelectEven() o1=a.SelectOdd() e1=e1.ColorMill(25700, 25700, 25697, 25688, 25700, 25700, 25700, 25700, 25700, 25700, 25700, 25700, 25700, 1124, 5) o1=o1.ColorMill(25700, 25695, 29794, 25700, 25711, 25700, 25700, 25700, 26468, 25700, 25700, 25700, 25700, 1124, 5) Interleave(e1,o1) # Re-assemble the frames in their original sequence #-----Color tweak + denoise chroma for all frames # ColorMill(25702, 25698, 25697, 24416, 25704, 25700, 29044, 25713, 25700, 25700, 25444, 25700, 25700, 1124, 2053) # CCD(20,1) ########################
Code:ConvertToRGB32(matrix="Rec601",interlaced=false) #-----Correct tint in in Even and Odd frames ("e1" and "o1") to make them more alike. # AssumeTFF() # a=last # e1=a.SelectEven() # o1=a.SelectOdd() # e1=e1.ColorMill(25700, 25700, 25697, 25688, 25700, 25700, 25700, 25700, 25700, 25700, 25700, 25700, 25700, 1124, 5) # o1=o1.ColorMill(25700, 25695, 29794, 25700, 25711, 25700, 25700, 25700, 26468, 25700, 25700, 25700, 25700, 1124, 5) # Interleave(e1,o1) # Re-assemble the frames in their original sequence #-----Color tweak + denoise chroma for all frames ColorMill(25702, 25698, 25697, 24416, 25704, 25700, 29044, 25713, 25700, 25700, 25444, 25700, 25700, 1124, 2053) # CCD(20,1) ########################
I've looked, but cannot find a tutorial/documentation on using ColorMill. I don't do real well learning basics by trial and error; once the basics are mastered, experimenting for advanced usage comes naturally - it's just those first steps that trouble me. Probably have a need to get comfortable in the "surroundings". OCD? Anyhow, did a backward copy/paste in specs to see the script's ColorMill values when run under vdub, but haven't much of a clue why selected. The "middle point", "dark", "middle", "light" and "sharp/smooth" areas are unknowns. I do see that changing values changes the image, but don't understand the purpose of the various sections. Is there something similar that does have documentation? Saw another post where LordSmurf didn't think too much of color correction with either vdub or avisynth. Think sanlyn uses AE sometimes. My needs are probably not that advanced. I'd really like to adjust using histograms - I've a terrible eye for color matching & balance. Saw some tutorial stuff for ColorTools that made sense to me, but sadly, not Win7 usable.
Some of the older sites use the flaXen filter for vhs a lot, but I'm not seeing now. Been replaced in usage?
sanlyn, have all the filters that have parameters used in this script been tweaked specifically for the clip or are some pretty much "standard" for vhs?
(ccd_sse2) It once had a text help file, but it was in Russian and the text is corrupt, so not even Google Translate could handle it. Translate just returns the same binary gibberish. Phooey. Fortunately the settings are simple.
Code:Camcorder color denoise 1.6 MT The filter is designed to remove from the interlaced color noise matrix cameras. The default setting is optimized for camera canon md160. Applying a filter in AviSynth: LoadVirtualDubPlugin ("c: \ Program Files \ VirtualDub \ plugins \ ccd.vdf", "ccd", 0) ConvertToRGB32 (interlaced = true) ccd (30,1) where: 30 - level of treatment, 0-100 (over 100 works, but undocumented) 1 - includes multithreading, 0 - off The number of threads is determined automatically depending on the type of processor, up to 4.
Last edited by dianedebuda; 3rd Mar 2014 at 19:01.
-
Another reason why I don't use AvsPMod. Whenever you get errors of this kind, tell us precisely what the error message says.
What crash message? If you mean the generic "Ooops! VirtualDub has crashed", it could mean anything. Hopefully you have all 32-bit apps and filters. You can't mix 32-bit and 64-bit components.
I don't know that the split script worked or not. If you split the script into 2 parts and the second part begins with the ConvertToRGB line and the ColorMill, SelectEven/Odd, etc., lines follow it, you should have "return last" as the last executable statement in the #2 script.
CamcorderColorDenoise (CCD): it's to remove chroma noise (obviously). True, the original text "help" is no help. I usually set the filter strength by finding some chroma noise and eyeballing the results. If you don't know what chroma noise looks like, there are two website articles that discuss it in terms of digital cameras, but tape has chroma noise as well and both sources can look very similar:
http://www.cambridgeincolour.com/tutorials/image-noise.htm
http://www.cambridgeincolour.com/tutorials/image-noise-2.htm
If you want to see some specific chroma noise in the guitar video referred to, look at the shadow details, such as the shadow side of the man's guitar, his slacks, the audience, etc. You can see it better if you brighten the image so that shadow noise is more visible.
There are lots of histogram articles. Most of them deal with still photos, but the color principles are the same for stills and video. Two articles that I saw on RGB a long long long time ago are bstill around. I found them helpful:
http://www.sphoto.com/techinfo/histograms/histograms.htm
http://www.sphoto.com/techinfo/histograms/histograms2.htm
A histogram doesn't give ylou all th answers. You have to learn how to spot problems by eye as well. What do you "compare" the colorws to? You compare them to the eay the world looks outside of a video. People with blonde hair should have blonde hair, not green or purple. A white dress ought nto be white, not tinted orange. A black suit should look balck, not some other color. In itself, a histogram might not tell you if the whites aren't white, although sometimes the shape of the histogram at the bright end will give you a clue about what's wrong.
The tutorials for the gradation curves filter also bring up some good points:
http://members.chello.at/nagiller/vdub/tutorial/tutorial.html
There are no standards. Every video is different, regardless of source. Sometimes others and myself set up filters for a single scene, sometimes several scenes, sometimes a setting will work for most of a video. VHS in particular is extremely inconsistent from minute to minute. If autowhite is in a camera (as it is in the clips I've seen so far in this thread), that's another problem.
MT settings: I lost patience with MT. Many people don't use 2.6, so including MT settings is beside the point in posted scripts unless the issue is being addressed specifically. For me, it takes less time to let a script run slow in Avisynth 2.5 than it does to try re-inventing the wheel with MT . That's just me. Sometimes MT is so ornery I'd end up running 2 or 3 separate scripts. Hassle.Last edited by sanlyn; 19th Mar 2014 at 03:20.
-
Flaxen's VHS filter is an oldie. I seldom use it, and omnly for shifting colors a few pixels left, right, up, down. I turn the other features off.
Most of what you learn about ColorMill is just from using it, along with principles picked up from using other color filters.
"Sharp/Smooth): Don't use it. A useless holdover from earlier versions. There are better sharpeners.
Dark/middle/Light: Dark means, of course, dark colors and luma values. "Light" means light colors, and "middle" means everything in between. The ranges don't work with sharp cutoff points, but transition gradually between the three ranges. Similiar to the color wheels you see in other apps, but it uses sliders. If you want to identify the operational ranges, think in terms of splitting the histogram into three parts: the darks include RGB 64 on down, brights inclkude RGB 192 and up, and the middle is centered at RGB 128.
Middle Point: This is often a useful tool, and it's usually one of the last tweaks applied. It's a little like a gamma adjustment, which raises or lowers midtones in the RGB 128 range, which is right in the middle of the color spectrum from 0 to 255. Once you understand that RGB 128 luma and color values are in the "middle", you can understand what "Middle Point" does: it shifts the middle point from 128 up or down in RGB value. If you raise the middle point, the midpoint is adjusted from 128 to a higher spot in the histogram. Lower it, and the processing considers the middle point to be lower than 128. I'll demonstrate this shortly.
Okay, folks a little color theory. Can't give a complete course here (you'd fall asleep, and so would everyone else), but...
In video there are three primary colors. You probably know what they look like. Usually a color patch for each primary is shown at the brightest RGB values, and the RGB values are stated (usually) are stated as "RGB -999-999-999) , the 9'd being replaced with the values for Red, Green, and Blue, in that order. Sometims you just want to specify a range, such "up there where it's bright, around RGB 220). RGB 220 isn't a color, it's a spot in the RGB spectrum. If you want to state a specific color, give all 3 RGB values. A question everyone asks is, why do I have to know the primary colors? Because all hues are based on the primaries, along with their brightness values. Unlike YUV, which stores luma and chroma info separately, RGB stores luma and chroma info in the same pixel, not separately.
The primary colors are:
[Attachment 23873 - Click to enlarge]
You can see that each primary is "pure", i.e, it doesn't have elements of the other two colors mixed in with it. There are three more colors, all based on the primaries. Should you know what these "secondary" colors are? Yep. You'll soon see why. You get the three secondary colors by mixing two of the primaries. The three secondaries are created by mixing two primaries. The secondary colors are:
[Attachment 23874 - Click to enlarge]
Here is why the secondaries are handy. When most people say "my video is too red", what do they do? They add blue. Tsk, tsk. Mistake. You correct red by adding its opposite, which isn't blue. Adding blue to red gives you pink. If a skin color is too red, adding blue seldom makes the skin look natural; it makes it look too pink. The color that is "opposite" to red in cyan (blue + green), not blue. Here are the general rules:
- To correct too much red, either reduce red -or- add cyan (green + blue).
- To correct too much green, reduce green -or- add magenta (red + blue)
- To correct too much blue, reduce blue -or- add yellow (red + green).
- To correct too much grenish-bluish cyan, either reduce cyan -or- add red
- To correct too much pinky-purply-magenta, either reduce magenta or add green
- Too correct too much orangey/pumpkin/yellow, either reduce yellow or add blue
To correct too much red you might need just a little green and a lot of blue, or vice-versa, or you might need equal parts of blue+green, etc. How do you know how little or how much? It depends on how far off the color is and how you want the result to look. It also depends on knowing the pixel values in the color you're trying to correct. How do you know what the pixel values are? You use a pixel value reader. Sometimes you can tell by just looking, but you'll make twelve eyeball guesses before you get it right. Use a pixel reader. All the big-time apps have pixel readers. For a reason. Even Vegas Pro has one. Sometimes you can tell by looking at an RGB histogram, sometimes not.
So you've learned six colors. But there are really 9. The other three are three different shades of "white". Should you know about these as well? Yep. You'll see why. The other three "colors" are:
[Attachment 23875 - Click to enlarge]
Notice something about the RGB values of these three colors. In each case, the "color" consists of equal proportions of all three RGB primaries. You get different shades of white, light grays, middle gray, and dark grays by mixing equal portions of each primary. Dark gray is RGB 64-64-64. Black is all zeros, meaning "no color". Why should you know this? If you have an image in which the whites don't look white, the grays don't look neutral gray, and blacks have a tint, it means that the primaries are out of whack. You have too much of one, two little of two others, and so forth. None of the other colors will look "correct".
Photogs and colorists set color balance at three points, if possible: white balance, gray balance, and black balance. If you have an image in which you have some black stuff (or nearly black), some neutral grays, and some light or bright white, and you can make them look "correct" by adjusting primaries at dark, middle, and white ranges, then all three primaries will be in balance and those colors and the other colors will fall into place. True, you can't make any one of those hues look exactly right. But you can get close.
What if you don't have anything white, gray, or black in the image? You are up a tree, and you have to count on experience and two good eyes. Sometimkes you have some odd scene, such a night club or theatrical stage shot with a lot of colored lights; those can be a problem, and you have to look around for clues and just fight it out. Experience, of course, goes a long way here. But knowing some color theory, reading pixel values, and consulting a histogram can help quite a bit.
An example of how mixing three primaries results in some commonly seen colors:
[Attachment 23876 - Click to enlarge]
Other essentials: Don't try to correct color with the sun from a brightly lit window shining on your monitor. Work in subdued light. Don't place a lamp directly in front of your monitor. On the other hand, don't try to work in darkness -- your eyeballs will shut down, thinking the monitor is too bright, and besides that you can't see your keyboard. Don't stare at a scene continuously for two hours: your brain will adjust to whatever color and brightness values are in front of you, and your brain will try to make it look "right'. Turn away,look at something else after 20 minutes or so, and come back. But don't stare at a bright light during your break,or it will take several minutes for your eyes to open up again.
Second essential: use a calibrated monitor, even if you have to use some of the funky online free calibration patches. The typical PC monitor is a big fat liar. Out of the box they have horrible color imbalance, murky shadow detail, and they're too bright.
Try the color patch test below. It has six "RGB 16" color patches. Actually, only one of them is RGB 16-16-16 "black". The other five are off-color. The best way to know for sure is to use a pixel reader. The large dark gray background frame is RGB 64-64-64.
[Attachment 23878 - Click to enlarge]
There are two columns of black patches. The left-hand column is patches numbered top-to-bottom as 1, 2, 3. The right-hand column is numbered top-to-bottom as 4, 5, 6.
Will be back later to demonstrate how these principles can be used to set levels and color balance in the guitar clip using ColorMill and gradation curves. Please excuse any stupid typos in this post. I'll correct them later. Gotta go fix dinner.Last edited by sanlyn; 19th Mar 2014 at 03:21.
-
On now to using some of the above principles for working with ColorMill. I'll use the same guitar practise video, but without denoising or prep with ColorYUV or other Avisynth filters. I opened the guitar clip directly in VirtualDub and loaded the ColorTools histogram. Frame 232, unprocessed, looks like this in RGB:
[Attachment 23883 - Click to enlarge]
First thing the histogram tells you is that black levels in the image don't go below RGB32 (dark gray), and the only blacks are the RGB 16 borders. Color isn't too bad and shouldn't require major surgery, although blue starts dying out by RGB 180 or so. As the clip progresses and the camera zooms in, there's also a color balance change in the darks from slightly warmish but tolerably clean-colored darks that take an unfortunate turn toward reddish purple, thanks to the auto camera features. There's some slight shadow detail loss and a little bright clipping in-camera (it's a seriously wide contrast range that you're asking autogain to handle). Brighter areas of skin tones are a little yellowy or greenish but not too bad considering the mixed lighting. Whites are about 25 RGB points deficient in Blue compared to Red and Green, so whites aren't really white. Red is oversaturated, but I'd prefer to let FixChromaBleeding help solve that problem first. It's in the original script, but isn't used here.
As with ColorYUV, the first step is to correct levels and make use of more real estate in the histogram. ColorMill is used first to get better levels and a little "zip" into the image. Below, the "Levels" function of ColorMill lowers the blacks from about RGB 64 on doiwn, mwith most of its effects on the very dark colors. Levels is then used to perks up the brights above RGB 80 or so, with most of the effects in the upper RGB 200's. Have to take care here to avoid overshooting the 16-240 limits -- which is often too easy to do with ColorMill. I decided that the midrange looks OK, so the 'middle' slider stays at default center position. The settings for the LEVELS sliders are shown in the image below, left-hand side.
[Attachment 23884 - Click to enlarge]
(above, right) The next step is a little color correction at the bright end. In the image of frame 232 above, notice the young girl in the right-side background. She's wearing a white blouse-- well, it appears that it's supposed be white. It's apparently summer season and it's a thin blouse that shows darker white or other clothing underneath, but the collars and button panel are the brighter parts. Rather than white, the original after levels are corrected gives typical pixel readings of RGB 230-220-210. We wouldn't want to make that a really sparkling-fresh white, but certainly not a reddish yellow despite the warmish lighting. In the image above, the "RED-GREEEN-BLUE" panel and the "Light" settings for bright colors are used to lower bright Red, raise Green slightly, and juice up that deficient bright Blue. This gets the brightest white at RGB averages of 227, 224, 220, which looks more natural in clothing than screaming bright white does.
The results of the ColorMill changes, below.
[Attachment 23886 - Click to enlarge]
Note that YUV prep wasn't used here -- not a total necessity, really, but denoisers and chroma fixes in Avisyth can affect color balance, especially if a lot of chroma noise is cleaned. In this case I thought it better to show ColorMill processing only. Also, the ContrastMask in the original script gave somewhat better shadow detail. And note that some of the reflection off the boy's guitar and some other areas are "hot spots" of bright color. It's sometimes difficult to avoid with many color filters. But that can be fixed by chaining other filters.
I believe slight saturation increase was applied with ColorMill in the original script. This demo doesn't use anything like the settings in that script, but here is a demo of ugly oversaturation (which, as a matter of fact, looks "great" to a lot of users -- me excluded). As shown in the panel below, the "SATURATION" panel in engaged, and the "stick" button is activated. The "stick
buttons locks all the sliders in tandem, so moving one slider moves all of them together. This affects saturation in the darks, middles, and brights. (The "Lock" button locks your settings so that you don't accidentally change them. But Lock isn't used here).
[Attachment 23887 - Click to enlarge]
Below, a demo of the the MIDDLE POINT function, but not one that you would want to use on this clip. In this case, the Middle Point is raised well above RGB 180 or 190. Notice that the brights and midrange are looking rather "raw" and overexposed, but the darks don't really suffer very much. The Booster and the Base Shift sliders (not used here) do two things: the first one modifies the strength and range of black levels below the raised midpoint, and the second slider applies a similar mod to the bright end. They are seldom used, but I've tried them in the past and they do work.
[Attachment 23888 - Click to enlarge]Last edited by sanlyn; 19th Mar 2014 at 03:22.
-
I've seen users say that it might help to run Virtualdub in Compatibility Mode. A Microsoft page discusses this. http://windows.microsoft.com/en-us/windows/make-older-programs-run#1TC=windows-7.Odd web page, they waity until the bottom of the page to give you this quick tip:
Tip
You can also open the Program Compatibility troubleshooter by right-clicking a program's icon or shortcut and then clicking Troubleshoot compatibility.
I'm afraid Microsoft and even Mac have little interest in supporting video processing for anything but PC display for the last few years. That's why many users rebuild or build XP machines.Last edited by sanlyn; 19th Mar 2014 at 03:22.
-
I don't use AvsPMod. Whenever you get errors of this kind, tell us precisely what the error message says.What crash message? If you mean the generic "Ooops! VirtualDub has crashed", it could mean anything. Hopefully you have all 32-bit apps and filters. You can't mix 32-bit and 64-bit components.
I don't know that the split script worked or not. If you split the script into 2 parts and the second part begins with the ConvertToRGB line and the ColorMill, SelectEven/Odd, etc., lines follow it, you should have "return last" as the last executable statement in the #2 script.
What I didn't know is if the file output by vdub after Interleave was good since it immediately crashes. Don't know if error occurs after file output is complete or msg display is delayed until file done. File plays, but is there a way for me to verify that the fields are back in order? I have a test file using the same input with just the ColorMill color correction (what followed the even/odd processing in the orig script) and the file size matches the Interleave one. Avisynth info matches for both of them too. Maybe I could do some kind of funky color cast to the even fields after they are separated & see if they show up properly mixed in the output after Interleave?
CamcorderColorDenoise (CCD)
There are no standards.
MT settings: I lost patience with MT.
Flaxen's VHS filter is an oldie
ColorMill Dark/middle/Light: Dark means, of course, dark colors and luma values. "Light" means light colors, and "middle" means everything in between.
I've seen users say that it might help to run Virtualdub in Compatibility Mode
Again, many thanks, sanlyn. -
Hmm. That WIn7 is a problem. The PC I just built has dual-boot with XP and Win7. Guess which one I'm using all the time? (XP, it seems). I have too much good software and two good ATI cards to leave XP behind. I know several people who feel the same way.
I have Avisynth 2.5 and 2.6 dll's. 1 out of 3 times I use 2.6, something goes haywire. I'll keep at it, but when I want to get something done I copy 2.5 back into System32.
Main reason for the routine that makes even/odd frames more alike: without correction, the color is more difficult to clean up, especially when the color balance changes during that camera zoom. But mainly there's a vague flicker during playback that won't go away. Anti-flicker and other strong filters didn't help.Last edited by sanlyn; 19th Mar 2014 at 03:22.
-
Not sure if I should PM dianedebuda for that, but my guess is others may be interested :
@dianedebuda :
As you maybe noted, Sanlyn deleted all his posts, including in this thread, the attached scripts.
By any chance, could you upload them ?
(if you still have them, so I (or others) could benefit from it, following this quite useful discussion about VCR capture and all the luma/chroma problems.)
Thanks -
Wow - all of sanlyn's post are gone & he's listed as Banned. Shocker. What a loss. Wonder what happened. Been busy with some other stuff, so haven't been around for several months. Have most of the scripts & will try to find time to repost in a few days.
DianeLast edited by dianedebuda; 25th Jun 2014 at 13:28.
-
I saved some, too. Not on my home PC for a couple of days, though. Will take a look later.
- My sister Ann's brother -
Apparently, a thread discussion about HDMI cables went out of proportions, and, as some videohelpers didn't agree (at all) with Sanlyn's pov, he felt nobody was trying to understand him or his vision about what makes "image quality" in a not-purely technical way.
And so he deleted all his VideoHelp's posts, including all the ones in others threads.
Seems silly, and it's a shame, indeed.
But if you guys still have some scripts, well that's still "some" consolation.
Thanks.
Similar Threads
-
Hi8 Capture problem
By manyon in forum CapturingReplies: 2Last Post: 24th Nov 2013, 18:05 -
Hi8 capture- Suggestions for Harddrives and Capture Cards?
By CMP1223 in forum CapturingReplies: 7Last Post: 10th Jun 2011, 23:05 -
Should I capture VCR, 8mm and Hi8 tapes to MPEG2 or MPEG4?
By dorenhagen in forum CapturingReplies: 18Last Post: 25th Jan 2011, 22:00 -
How to evaluate your work
By persephone in forum Newbie / General discussionsReplies: 5Last Post: 17th Jan 2011, 11:42 -
Hi8 VCR help
By Famicoman in forum Newbie / General discussionsReplies: 1Last Post: 27th Oct 2010, 18:32