Problems With QTGMC Need Help

[real.finder]

You dont need AnimeIVTC or even smdegrain required plugins for qtgmc you just need them loaded for some shared functions in them

Are you really made smdegrain in .avsi not .avs? If you made it .avs it will not auto loaded from plugins folder

[real.finder]

I told him before that he should convert to YV12.

Why yv12? yv16 and yv24 or even HBD are possable since age. But anyway for yuy2 yv16 is same color subsample

[jagabo]

I told him before that he should convert to YV12.

Why yv12? yv16 and yv24 or even HBD are possable since age. But anyway for yuy2 yv16 is same color subsample

Because YV12 has been supported since day one. And he's probably going to end up with YV12 in his final product anyway. He should start with the most widely supported and work his way up later if necessary.

[dellsam34]

Error message for line 494 is caused by a wrong extension of the smdegrain file, I had .avs instead of .avsi, Finally no more error messages.
Vdub is rendering the video at 5 frames a second is this normal or my CPU is slow?

[dellsam34]

I was not impressed by my first de-interlacing job, a lot of motion blur, although I'm just using the basic QTGMC() commend, How do I really experiment with this since I can't monitor video in real time, What commends should I be looking for?

[jagabo]

Vdub is rendering the video at 5 frames a second is this normal or my CPU is slow?

Since you are using AviSynth+, use the Prefetch() command at the end of your script:

AviSource("filename.avi")
QTGMC()
prefetch(6) # about 3x the number of cores.

I use 12 on my quad core i5 2500K CPU and get about 35 fps with your sample (not including encoding time). You can also use faster settings, like preset="fast" (70+ fps on my system), slightly lower quality (Still shitloads better than any other deinterlacer).

I was not impressed by my first de-interlacing job, a lot of motion blur, although I'm just using the basic QTGMC() commend,

I gave you one example in your "yadif" thread. And read the docs. I never see motion blur, just a little loss of fine detail from noise reduction. As I mentioned in your other thread there is evidence in your sample that the video is a PAL to NTSC conversion with blended fields. You might interpret that as motion blur. Upload some samples so others can see what the problem is.

How do I really experiment with this since I can't monitor video in real time

Work with short clips. Encode, compare, change settings, encode, compare, change settings... Or use an AviSynth script to compare two videos side by side:

AviSource("filename.avi")
StackHorizontal(Yadif(mode=1), QTGMC())

v1 = AviSource("encoding1.avi")
v2 = AviSource("encoding2.avi")
StackHorizontal(v1, v2)

Or interleave frames (for flipping back and forth in an editor):

v1 = AviSource("encoding1.avi")
v2 = AviSource("encoding2.avi")
Interleave(v1, v2)

[dellsam34]

The tape captured is NTSC and captured in NTSC I don't know how it behaved like a PAL, maybe because the documentary was shot by a film camera and got transferred to NTSC with pull down of some sort.

I added prefetch(6) and that added a couple of frames a second, So my computer is slow.

[dellsam34]

I'm still liking Vdub/Yadif over QTGMC:

Here is the original sample

Vdub/Yadif sample

Basic QTGMC sample

Maybe adding some tweaks makes it better but overall Yadif looks better and works faster the files are bigger however.

[manono]

The tape captured is NTSC and captured in NTSC...

Are you writing of the same video you uploaded in your next post? The one named Night Gathering.avi from "Here is the original sample"?

Because it's PAL all the way, 720x576 and 25fps. It has bad edge enhancement (aka sharpening halos) probably done by you. Have you made sure to turn off all sharpening in the menu or wherever the sharpening settings are found?

Yes, Yadif is faster. Much faster. But it's nowhere close to being better than QTGMC. But at least after all this you stuck with it and got QTGMC going, if only to decide you didn't like it much.

[jagabo]

I'm still liking Vdub/Yadif over QTGMC:

Here is the original sample

Vdub/Yadif sample

Basic QTGMC sample

Regarding deinterlacing quality: look at the near horizontal edges of the door frame and picture on the wall. Notice how they're buzzing in the Yadif video, no buzzing in the QTGMC video. If you look at still frames you'll see bad aliasing on those edges with Yadif, nice and smooth with QTGMC.

Maybe adding some tweaks makes it better but overall Yadif looks better and works faster the files are bigger however.

You have the wrong field order in your QTGMC encoding. Add AssumeTFF() before calling QTGMC (AviSynth assumes BFF when the field order isn't specified). I think the fast jerkiness is what you're calling motion blur.

[dellsam34]

Are you writing of the same video you uploaded in your next post? The one named Night Gathering.avi from "Here is the original sample"?

Because it's PAL all the way, 720x576 and 25fps. It has bad edge enhancement (aka sharpening halos) probably done by you. Have you made sure to turn off all sharpening in the menu or wherever the sharpening settings are found?

Yes, Yadif is faster. Much faster. But it's nowhere close to being better than QTGMC. But at least after all this you stuck with it and got QTGMC going, if only to decide you didn't like it much.

This is a PAL home video, the other sample is a prerecorded tape NTSC, Sorry for the confusion. I haven't done any enhancement on any video, Just lossless capture.

[dellsam34]

You have the wrong field order in your QTGMC encoding. Add AssumeTFF() before calling QTGMC (AviSynth assumes BFF when the field order isn't specified). I think the fast jerkiness is what you're calling motion blur.

I see, I will give it a shot.
On the PAL video is looking like jerkiness, But other NTSC samples it's more of a motion blur than jerkiness.

Edit:

Yes the field order was wrong, It looks better now.

[manono]

I haven't done any enhancement on any video, Just lossless capture.

While some VCRs have built-in sharpening that can't be modified, most have some sort of sharpening controls that might be on strongly by default. Have you looked? If it's adjustable, it's way better to get rid of it before capping, rather than trying to remove it later on.

[dellsam34]

Yes my VCR has a picture mode, Sharp, Soft and Edit, I always try the best option before started capturing. Also I can turn off and on DNR from the menu and TBC from the front panel, I'm always checking those not only for every tape but for every scene on the tape.

[manono]

Sounds like a good VCR. So, the sharpening is always set to Soft (unless Edit means no sharpening), and DNR is always turned off?

[dellsam34]

It's set to whatever you leave it on, I leave the DNR almost always ON unless it gives me problems, it cleans up the chroma noise such as blue and red grains, TBC stores the scan lines in a digital buffer and readjust the timing so they are stacked evenly to reduce the comb effect that appears on the edge of the video frame. JVC makes good VCR's, they invented the damn thing.

[jagabo]

Generally, if you're planning on filtering after capture sharpening should be set to edit (no sharpening). DNR should usually be turned off. TBC should be enabled. And you should be capturing s-video, not composite.

[dellsam34]

I don't do any picture enhancement after capturing, it's time consuming and requires a lot of know how and in most cases it comes at a price of slightly sacrificing the picture quality, The DNR on the VCR works beautifully at the source (JVC engineers knew what they were doing) I never had a problem with it so far so there is no reason for me to capture with noise and try to remove it later.

[jagabo]

I don't do any picture enhancement after capturing

Then what are you doing with QTGMC? Use whatever your VCR gives you and live with what you get.

[dellsam34]

I don't do any picture enhancement after capturing

Then what are you doing with QTGMC? Use whatever your VCR gives you and live with what you get.

Just de-interlacing.

[dellsam34]

I'm at a loss here, I'm trying to de-interlace this VHS home video from 1987 -sample attached- but I don't seem to really go beyond the basic QTGMC() commend, What should I be looking for?
Here is the original sample and the de-interlaced one.

[jagabo]

What are you looking to change?

[dellsam34]

There is motion artifacts in fast scenes, Does it look ok on your end? I tried VLC and MPC-HC, MPC-HC is little better than VLC. Also if I crop the unwanted edges with QTGMC does it replace the cropped lines by doubling from the existing ones throughout the frame or it just cuts off the edges and lower the resolution?

[jagabo]

I don't see motion artifacts. The original video has motion blur during fast motion. No deinterlacer is going to remove that. There is some noise reduction which you can reduce with TR2=0. As a beginner you can try the presets: QTGMC(preset="fast"). The presets are: "Placebo", "Very Slow", "Slower", "Slow", "Medium", "Fast", "Faster", "Very Fast", "Super Fast", "Ultra Fast", "Draft". The default is "Slower".

There are lots of settings described in the docs. Try QTGMC(SourceMatch=3, Lossless=2, Sharpness=0.1, TR2=0)

[dellsam34]

Is it possible to crop 6 lines from the frame and assign aspect ratio with QTGMC? The reason for cropping is because my captures are 720x486 hardware locked.

[jagabo]

You should crop interlaced 4:2:2 video by multiples of 2 to at the top to avoid reversing the field order. 4:2:0 video should be cropped by multiples of 4 at the top to avoid reversing field order and blurring the chroma.

480 and 486 line NTSC caps have the same sampling aspect ratio.

[dellsam34]

I think I figured out the cropping part, Since I'm converting my caps to HuffYUV using Vdub anyway I may as well do the cropping right there, I trim off 2 lines from the top and 4 lines from the bottom which gives exactly 720x480 non square pixel, Then I do the de-interlacing with QTGMC, the only thing left to do is the aspect ratio flag, If QTGMC doesn't have an option to do this I will have to do it during the compression to mp4, now I need a new commend for ffmpeg to convert from a de-interlaced video, I use to do it from an interlaced video using the following command:

ffmpeg -i input.mov -vf "crop=w=720:h=480:x=0:y=2,scale=w=-1:h=-1:inte rl=1,format=yuv420p,setsar=sar=8/9" -flags +ildct+ilme -c:v libx264 -crf 10 -x26 4opts bff=1:colorprim=smpte170m:transfer=smpte170m:colormatrix=smpte170m:force- cfr -c:a aac -b:a 192k Output.mp4

What would have to change in that command to address the issue? I know I have to remove the cropping code for sure.

[juli4561]

Hi, im new in this world, today i try to upscaled a video but this was the error
[Attachment 55493 - Click to enlarge] how can i fix?

[jagabo]

Try removing the crop(). Does it work? What is the frame size of your source?

[juli4561]

Hi it doesnt work, i post another question here https://forum.videohelp.com/threads/399202-QTGMC-Avisynth#post2598316