Highest Quality Video Editing

[lomaidala]

VirtualDub Error
Avisynth open failure:
Not an Avisynth 2 C Plugin: C:\AviSynth Plugins\masktools2-25.dll
(C:\Movie.avs, line 3)

[manono]

Yadif is a C Plugin but Masktools2 isn't. Remove the 'C' from its load-plugin line.

[lomaidala]

Thank you. After comparing my new AVI, I am noticing that some lines (the drawn contours of the animation) appear to be broken, dotted, or flashing, compared to the way they used to look. Is there something from Yadif or Srestore that would cause this?

[manono]

Is there something from Yadif or Srestore that would cause this?

Yadif, yes. It can cause diagonal lines to become sort of jagged looking or thin diagonals to get gaps and flash/flicker/alias as you're describing. Which is why, in my original script, I wrote:

Yadif(Mode=1)#or the better QTGMC

QTGMC is much better, especially for the problem you're describing. But it's much harder to get working and it's much slower at the best settings. Most people think it's well worth the effort.

[lomaidala]

Thanks! I've decided to give QTGMC a try. I've installed the required plugins and Windows files, but I'm getting this error when I try to open my script:

VirtualDub Error
Avisynth open failure:
Script error: there is no function named "RemoveGrain"
(QTGMC-3.32.avsi, line 756)
(QTGMC-3.32.avsi, line 781)
(QTGMC-3.32.avsi, line 386)
(C:\Movie.avs, line 7)

[manono]

You have the wrong version of RemoveGrain. I don't know which it is and you might have to try them all. Anyway, try the one here:

http://avisynth.nl/index.php/Srestore

[lomaidala]

Thank you! I am really impressed with how good the video looks using QTGMC. It's great! Even on the "Slow" preset.

For anyone who wants to know, I ended up having to replace RemoveGrainSSE2.dll with RemoveGrainS.dll.

It'll be changed to 23.976fps, yes. You change it because that was the 'original' or 'base' framerate, except maybe for that fast moving intro part. If the intent is to create another DVD it'll still be better unblended and then encoded as progressive 23.976fps with 3:2 pulldown added, something the idiot DVD makers should have done to begin with. For anything else you want just the unblended and progressive 23.976fps video.

There is something with Srestore I'm not quite clear on. Sometimes a frame will get dropped entirely. I assume this is the nature of restoring the video to 23.976fps from 29.97fps. But it makes me wonder: why not keep the video as 29.97fps if it leaves more of the original animation intact? I can see what I am losing, but what am I gaining by restoring the video to 23.976fps?

[manono]

Sometimes a frame will get dropped entirely

If a unique frame does get dropped, then you can thank the idiot DVD makers for that. It never should have been field-blended to begin with.

But I'm not sure how you're telling a frame has been dropped. Again, though, the original framerate was meant to be 23.976fps (or 24fps). If one frame was dropped another must have been duplicated. If you upload a sample where that happens maybe we could get to the bottom of it (or maybe not).

However, with animations there are sometimes parts, especially during intros, that were drawn as 29.97fps. Generally, though, the vast majority of it is supposed to be 23.976fps and that's why you unblend it to that framerate. The alternative is to make a VFR video and I, for one, find that unappealing.

[jagabo]

There is something with Srestore I'm not quite clear on. Sometimes a frame will get dropped entirely. I assume this is the nature of restoring the video to 23.976fps from 29.97fps.

When SRestore() is working as intended no unique frames are dropped, only frames that are a blend of two other frames. So if blended frames were added to the original video to convert from 23.976 fps to 29.97 fps SRestore() can be used to remove those blended frames. If unique frames are removed you either picked too low a frame rate or there is something more wrong with the source than SRestore() is designed to handle. Or SRestore() simply made a mistake (which happens occasionally).

[hello_hello]

Thank you! I am really impressed with how good the video looks using QTGMC. It's great! Even on the "Slow" preset.

For anyone who wants to know, I ended up having to replace RemoveGrainSSE2.dll with RemoveGrainS.dll.

That'll work fine, but for the record you probably have a missing dependency, which is why RemoveGrainSSE2.dll wouldn't work. Probably MSVCR80.DLL.

Instead of RemoveGrain, Repair and VerticalCleaner, you might want to try RgTools instead. It's a replacement for those plugins and will possibly be a little faster. It requires Microsoft Visual C++ 2012 Redistributable Package. If you put RgTools.dll in the Avisynth plugins folder, remove the others.

[lomaidala]

Attached are two samples of the same clip, one with Srestore, the other without.
Compare frames 54-64 of "Srestore OFF.avi" to frames 42-50 of "Srestore ON.avi."
What I found was that on frame 56 of "Srestore OFF.avi," the character clenches his teeth when he closes his mouth, a frame that is completely absent from "Srestore ON.avi."

Instead of RemoveGrain, Repair and VerticalCleaner, you might want to try RgTools instead.

Thanks for the tip! Would there be a difference in quality by trying your method? Or just the difference in speed you mentioned?

[manono]

Your source is an XviD that's already been blend deinterlaced? What's the script for this one?

SRestore is usually used on field-blended (interlaced) sources.

[lomaidala]

Your source is an XviD that's already been blend deinterlaced? What's the script for this one?

Sorry, I didn't realize you wanted the source. My source is DVD. Attached is a new sample, after using DGIndex and Mpg2Cut2. Is that the best way to upload a DVD sample?

Also attached are my scripts.

[manono]

Your scripts were very bad. Maybe read the SRestore doc to understand how to unblend something:

Usage
--------
code examples:

bobbed (or progressive) source

srestore()

Your sample sucks as there's not enough there and flapping mouths are not very good to decide how it should be treated. What you're looking for in an anime sample is something where there's steady movement every frame. A slow pan or scroll or maybe an action scene, but nothing too crazy. Also, don't upload ZIP files as that just adds another step we have to go through to get at the sample. M2Vs are allowed for upload. I'm not positive about the true framerate because of the sample provided but try this:

QTGMC()
Srestore()

[lomaidala]

Sorry about the zips, and thanks! After doing a quick comparison, the frame that was missing is now back. I guess the problem was in my script.

I noticed you didn't include SelectEven() in your script suggestion. Doesn't removing this double the framerate? I'm now ending up with 25fps, which strikes me as odd.

Attached is a duplicate of one of the samples I uploaded previously. Is this what you're looking for?

[manono]

Thanks. Based on your new sample, now I think the framerate should be 23.976fps. Like so:

QTGMC()
Srestore(Frate=23.976)

I noticed you didn't include SelectEven() in your script suggestion. Doesn't removing this double the framerate?

As I quoted earlier, SRestore needs a bobbed video so it can choose the 'clean' bobbed frames and get rid of the blended frames.

I'm now ending up with 25fps, which strikes me as odd.

The majority of field-blended videos result from creating an NTSC video using a PAL source, hence the 25fps. However, I don't think that's what is happening here. I think they used a film source. Only the shabbiest fly-by-night companies do this.

[lomaidala]

Thanks! Looks like that did the trick.

What is a bobbed frame? Is it simply one that has not been blended?

Also, what about the fast-moving opening sequence? If it is 29.97fps, I would inevitably be losing some frames by using Srestore on the main body of the episode, right? Would it be possible to have 29.97fps for the opening, and 23.976fps for the rest if the final product is going to be an AVI?

While you are able to point out specific technical details I never would have been able to identify, your criticism of my DVD source is no surprise. These DVDs are reputed among fans to be a terrible release. Unfortunately, it's the only accessible release of the original series there is.

[jagabo]

Bob refers to the practice of turning the two fields of a frame into two separate frames. It ranges from a simple bob where the two fields are separated into two half height frames then resized back to full height, to a slightly smarter bob where the new frames use both fields in parts of the frame that aren't moving then fall back to something like a simple bob for moving parts of the frame, to much smarter bobs that look at motion between multiple fields/frames and try to fill in the missing lines with those other fields/frames.

AVI files are constant frame rate so you can't mix frame rates within the file. Some other containers support variable frame rates and would allow that. MKV and MP4 for example.

[lomaidala]

Thank you very much everyone for all of your help! I really appreciate it.

[lomaidala]

For all of my episodes so far, I've been combining the opening sequence from one DVD source with the main episode from another DVD source. It's worked great for the first 7 episodes, but with the 8th, something strange is happening.

The look of the video is coming out much more pixelated than usual. It appears as if QTCMC is not working properly, but it could be something else for all I know. The video looks great if I only use one video source. It's only when combining the two that something goes wrong.

Any idea what's going wrong here?

[jagabo]

How are you combining the intro and episode? Within the AviSynth script? Concatenating files after encoding?

[lomaidala]

I'm combining them using an AviSynth script in VirtualDub.

[jagabo]

So you have something like:

intro = WhateverSource("file1.ext")
body = WhateverSource("file2.ext")
intro + body
Filters()

That shouldn't cause encoding problems. Maybe it's problems decoding the sources? With complex filtering you often get frame read out of order -- causing the decoder to screw up. Some source filter have settings that may help with that. ffVideoSource() has "seekmode=0" for example. Mpeg2Source() sometimes has problems with MPEG 2 video in MKV or MP4 files. Demuxing to a raw M2V stream usually takes care of that.

[lomaidala]

This is my script:

LoadPlugin("C:\DVD\DGDecode.dll")
clip1 = MPEG2Source("C:\DVD\08\VTS_01_1.d2v")
clip2 = MPEG2Source("C:\DVD\24B\VTS_01_1.d2v")
clip2.Trim(16,2711)FadeOut(19) + clip1.Trim(2697,42399)
LoadPlugin("C:\Program Files (x86)\AviSynth 2.5\plugins\mt_masktools-25.dll")
Import("C:\Program Files (x86)\AviSynth 2.5\plugins\Srestore.avsi")
QTGMC( Preset="Placebo" )
Srestore(Frate=23.976)

So should I use DGIndex to demux the video to an M2V? If so, then what? I tried editing the D2V file in a text editor and replacing "VTS_01_1.VOB" with "VTS_01_1.demuxed.m2v", but that didn't work.

[jagabo]

I don't recall seeing DgIndex and Mpeg2Source() screw up with VOB files unless they're damaged. But if you demux from VOB to M2V I think you have to build new D2V files from the M2V files. You can't just change the filename in the D2V file because the D2V file contains pointers to the start of each GOPs in the source (hence the "indexing"). Those pointers will point to the wrong place in the M2V.

One other possible problem if you're encoding in bitrate mode -- maybe you're just not using enough bitrate for this particular episode. The artifacts from this usually look different than those from decoding errors. With decoding errors you get miscolored blocks, parts of the picture in the wrong place, repeated frames. When there's not enough bitrate you get rough, blocky looking images with low detail but they're not discolored and picture elements aren't displaced.

[lomaidala]

I don't recall seeing DgIndex and Mpeg2Source() screw up with VOB files unless they're damaged. But if you demux from VOB to M2V I think you have to build new D2V files from the M2V files.

Okay, so for each DVD, I load the VOB into DGIndex, save and demux, then load the M2V into DGIndex and save. The result is that I have a new D2V file that points to the new M2V file. I'm still getting the same result.

One other possible problem if you're encoding in bitrate mode -- maybe you're just not using enough bitrate for this particular episode.

I'm using Xvid's Target Quantizer, set at 1.0. Attached are two screenshots of my new AVIs using M2V videos as source--one showing the result of using only one video source, and the other using two video sources. From what I can tell, neither QTGMC nor Srestore are working in the second example.

[manono]

Okay, so for each DVD, I load the VOB into DGIndex, save and demux, then load the M2V into DGIndex and save.

Ordinarily, no. You load the VOB(s) into DGIndex and File->Save Project File. No demuxing. You then use the D2V you made in your AviSynth script. Together with the DGDecode.dll and MPEG2Source, you'll open the DVD in VDub or wherever.

I'm using Xvid's Target Quantizer, set at 1.0.

Why on earth would you do that? It makes a bloated file for not much gain in quality over using quant 2.

Your second picture is a mess, all right, but I don't know enough to be able to say what caused it. You did make the D2V using 'Honor Pulldown Flags' in DGIndex, right?

[lomaidala]

You load the VOB(s) into DGIndex and File->Save Project File. No demuxing. You then use the D2V you made in your AviSynth script. Together with the DGDecode.dll and MPEG2Source, you'll open the DVD in VDub or wherever.

Except for the most recent time, that's almosy exactly what I've been doing. I say "almost" because the only difference is that my D2V was automatically generated by DGIndex. I have no idea how to make a D2V in an AviSynth script. Is that possible? But this last time, I did what I thought jagabo was recommending: making a demuxed M2V and having the D2V point to it.

I'm using Xvid's Target Quantizer, set at 1.0.

Why on earth would you do that? It makes a bloated file for not much gain in quality over using quant 2.

That's good to know, I didn't know quant 2 was the recommended setting. I've been using quant 1 because I want the video quality to be as high as possible.

You did make the D2V using 'Honor Pulldown Flags' in DGIndex, right?

Yes.

[manono]

I have no idea how to make a D2V in an AviSynth script.

I said use the D2V you made in an AviSynth script. Oh, I see. I'll reword it. Sorry. "Use the D2V you created in DGIndex in your AviSynth script. Together with the DGDecode.dll and MPEG2Source, you'll open the DVD in VDub or wherever."

Better?

[lomaidala]

Ha, much better!

Well, it sounds like I'm doing everything right. Is it possible this is a bug? Anyone have any other ideas?