My Premiere Pro exports give me little black borders on the top and bottom

[CZbwoi]

Hello, I basically quit for a few months trying to do all of this when it just wouldn't work out. Now I finally have the will to get back at doing this project and again I'm in a brain numbing dilemma. I inserted camcorder files into Vegas, extracted the audio as Wav. Did color correcting in Vegas, frameserved it out into AvsPmod to add QTGMC, resize it, and to add ccd.vdf.

Here is that script:

AVISource("k:\Untitled.avi")
AssumeTFF()
ConvertToYV12(interlaced=true, matrix="pc.601")
QTGMC()

Spline36Resize(854,480)
LoadVirtualDubPlugin("C:\Users\CZ\Desktop\Video Capture\VirtualDub-1.10.4\plugins32\ccd.vdf","ccd") 
ConvertToRGB32()
ccd(100,0) # threshold (0-100), multithreading (0=off. 1=on)
ConvertToYV12() # or back to whatever you started with

I open up MeGUI 2624, load up the avs script, load up the Wav file (Nero AAC encoding), click auto encode, select MP4, start.

I've done this 3 or 4 times, the processing rate is 0.31 FPS with an estimated time currently remaining of 3 days, 23 hours, 57 minutes and 05 seconds. It goes down to 2 days eventually. When I first saw this I thought the number in front meant hours, and I saw okay, that's manageable I guess. But it turns out it meant days. This is for a 30 minute video.

Does anyone have any idea? I have already uninstalled the MT Mode version of the program(s) last I recall because of a different bug, so now I'm left at an insane crossroad. Does anyone have any inclination?

[poisondeathray]

What computer specs ?

What encoding settings ? You can use faster encoding settings

QTGMC is slow in single threaded mode, you can use a faster preset . Default is "slower" which means.... slower

Probably QTGMC is the bottleneck. Look at your CPU usage. On a decently modern quad core it's probably <25% . MT can speed it up, but if you run single threaded, most people would run parallel processes (hard to do with single vegas frameserver) , or use a lossless intermediate first (ie break up into stages)

You can assess bottlenecks by checking the speed at each stage with avsmeter or vdub analysis pass. For example, if you comment everything out after QTGMC, use avsmeter on that to check the FPS. That is maximum FPS being fed to the encoder. But x264 needs to look at previous/future frames. It has a lookahead function on top of search / motion estimation functions. So it will be substantially slower than the input FPS, but looking at bottlenecks is important to assess where the slowdown is occurring and what you can do about it

[CZbwoi]

Computer: Gateway FX6800-01e , Windows 7 64 bit, i7 CPU, 11 GB of RAM.

It was set at...16 I think, but I just set it at the default of 23 before I tried another test. Do you have a specific set of encoder settings, that gives the original picture quality frameserved out of Vegas and scripted out into an avs while being efficiently fast?..if that makes any sense. Right now I'm noticing that the current projected filesize is 365mb when previously with my old settings it was around 1gb...I think I'd like to change it back to that and preserve all quality if what I just did worsened it.

QTGMC is slow in single threaded mode, you can use a faster preset . Default is "slower" which means.... slower

Yep, shows how much I was out of the game lol, forgot about that. Here's what I changed it to now:

QTGMC(preset="fast",Sharpen(0.08, MMX=True))

To be honest I don't even remember what that MMX statement means, I copied it all from another script I still had, but it looks good. '-_-

MT like I said, if you look at the previous pages, caused a bug for me that made it impossible to use so I had to scrap that idea.

or use a lossless intermediate first (ie break up into stages)

I still don't quite get this method unfortunately.

Right now here's where I am, with you reminding me to change QTGMC to fast, it's now at 11 hours completion time going at 2.74 FPS for a 30 minute file. It's leagues better than where I just was earlier, but it's still 11 hours for 30 minutes worth. Is this still normal or what else can be done to speed up a single encoding session? Here's my processes (it won't let me upload an image for some reason): https://forum.videohelp.com/images/imgfiles/YqVG6EB.png

[poisondeathray]

8-9x faster is significant, but it still seems a bit slower than it should be on that computer (which google says is a I7-920), and it says 56% CPU usage on the screenshot so you have some slack/idle cycles

By encoding settings, primarily I mean what preset for the x264 settings like "slower", "fast" etc... just like QTGMC, or any custom settings you've used . x264 encoding speed can be 10x faster or 50x slower depending on settings

Also what sorts of filters do you have going on in vegas ?

You need to so some of the speed tests mentioned earlier to diagnose bottlenecks. That way you can tweak settings, filters until you get an acceptable quality/speed tradeoff you're willing to accept

When I says lossless intermediate, I mean instead of frameserving out of vegas, use a physical lossless intermediate encoded from vegas. All vegas manipulations are now "baked" into intermediate and it's faster to encode or split up encodes (if you have 1 big video, then it might not be as helpful, but you can still encode segments and join after). You also no longer have the overhead of vegas or dmfs

[CZbwoi]

By encoding settings, primarily I mean what preset for the x264 settings like "slower", "fast" etc... just like QTGMC, or any custom settings you've used . x264 encoding speed can be 10x faster or 50x slower depending on settings

It was set at 23, targeting quality, and medium. Now I'm trying to change it to 16 (for the better quality) and at a faster preset, but these encoding modes changing what that does is confusing me. Const. Quantizer, ABR, Const. Quality...

Also what sorts of filters do you have going on in vegas ?

Only white balance and color corrector.

AVSMeter is giving me the same result as I got in MeGUI with that script. Now I removed QTGMC(preset="fast",Sharpen(0.08, MMX=True)) from the script and AVSMeter is giving me around 6.7 FPS with a 2 hour 15 min completion time. This is obviously the problem but it's *the* reason I'm using avs scripts now (alongside the ccd filter).


So now I need some guidance on where to go from here. I need the effects of QTGMC (how it doubles the framerate, I guess 'deinterlacing' the video for a smooth effect) on my old home videos, my old camcorder videos, etc. I also need that ccd.vdf filter to get rid of the camcorder yellow spots effect on the camcorder files. Is there a substitute for these two things in Vegas? Can I add plugins similar to these things in Vegas? If there isn't anything like the ccd.vdf one, I could frameserve out like I am now just for that...but for QTGMC-like effects is there anything that can be done in Vegas instead, as that's the bottleneck here? I don't know what else I can do, I feel like if I set the preset to Ultra Fast or Draft that something unpredictable will happen.

There has to be something Vegas can provide like it/them, right? Or does nothing compare?

In that way I can do all the magic in Vegas, export it as whatever and keep that file as the final, or afterwards shove that into MeGUI 2624 for further compressing.

[poisondeathray]

So now I need some guidance on where to go from here. I need the effects of QTGMC (how it doubles the framerate, I guess 'deinterlacing' the video for a smooth effect) on my old home videos, my old camcorder videos, etc. I also need that ccd.vdf filter to get rid of the camcorder yellow spots effect on the camcorder files. Is there a substitute for these two things in Vegas? Can I add plugins similar to these things in Vegas? If there isn't anything like the ccd.vdf one, I could frameserve out like I am now just for that...but for QTGMC-like effects is there anything that can be done in Vegas instead, as that's the bottleneck here? I don't know what else I can do, I feel like if I set the preset to Ultra Fast or Draft that something unpredictable will happen.

There has to be something Vegas can provide like it/them, right? Or does nothing compare?

In that way I can do all the magic in Vegas, export it as whatever and keep that file as the final, or afterwards shove that into MeGUI 2624 for further compressing.

You can double rate deinterlace in vegas, but the quality is quite a bit lower. QTGMC really shines here. Even on some of the faster settings it's much better. There are many comparisons out there. (Basically there is no substitute, quality wise; not even in other programs, like Adobe ,Nuke etc... This really is one filter that has nothing even close)

There are various denoise and 3rd party filters in vegas that can get rid of noise, yellow spots. Neat video for example.

YOU have to decide where you want to make trade offs for speed vs. quality . You can run some short tests if you like. For some scenarios/goals, maybe the lower deinterlacing quality is "adequate". For others, sometimes you want to do a better job and if takes 3-4x longer, so be it.

[CZbwoi]

I see, that makes sense. I was quite pleased with the super fast setting on QTGMC, seemed to be the same quality as before to be honest. But I'd like to tryout the ultra fast preset now and have no idea how to toggle it. It tells me that I need to insert yadif in the script as well, but I don't know how or where, I've tried implementing it a number of different ways but all of them give me an error. I'm looking in my plugins folder and yadif.dll is there, but I don't know how to activate it along with ultra fast QTGMC in my script. Can you give me the code line(s) I need to insert to make that work?

[poisondeathray]

yadif doesn't autoload like most avisynth .dll's. You need to load yadif.dll with Load_Stdcall_Plugin

Load_Stdcall_Plugin("PATH\yadif.dll")

where PATH is the actual path

My opinion - if you're using "ultra fast", you might as use something other than QTGMC. "super fast" is the fastest preset that still uses nnedi3 for the edimode . Once you get into yadif territory , it's probably not worth it

[CZbwoi]

Alright, sounds good and you know more than I, I'll probably stick with super fast then. Thanks.

I've also got some other questions back on the saturation and color correcting side of things, which is proving to be the most difficult choice to me before I start cranking out these final exports, but I think I'd be better off starting a new thread over that...I don't know.

[CZbwoi]

Hello PDR, was hoping you could help me out with a similar problem like this thread first began on, but now having to do with ccd.vdf and Avisynth. I don't remember if it was you that first pointed me to using ccd.vdf over here to get rid of my blue/yellow spots.

I'm frameserving out of Vegas into Avisynth, and when I add an instance of ccd.vdf that I need, it adds these black little inside borders on half the picture. Regardless of the value(s) entered. I went to a frame with a white background so it would be viewed more easily.

Original:




ccd.vdf added:



I don't understand why this is happening and it doesn't make any sense. As you can see, a thin black inside border is added along the top of the picture. Then overlapping it on the right hand side is an upside down L shape thicker black line. I don't even know what could cause this. And yes, they are present in the final exported files as well exactly as displayed.

Neat Video would be another learning stone for me and I just want to get the effect that ccd.vdf's "25" percent value gives me, exactly that, and now it's all here and perfect but these black borders appear from out of nowhere. Ever seen anything like this?

[poisondeathray]

Not sure, I don't really use ccd .

But here are some observations:

But if I take a random 854x480 video, it pads to 856x480 afterwards if you notice the dimensions. Some filters require mod 8 or mod 16 dimensions, but usually not ones that work in RGB, usually only ones that work in 4:2:0 subsampled YUV . But I can replicate what you are seeing with a 854x480 input. A 848x480 (mod16) input doesn't exhibit that issue . So you might move the Spline36Resize(854,480) step later, after ccd , since you were starting with a mod16 video from vegas IIRC

[jagabo]

ccd messes up with mod 2 frame sizes. Change it to 852x480 (mod 4) or 856x480 (mod 8) and the problem goes away.

[SameSelf]

I hate to jump in and potentially confuse the OP, but why haven't you configured QTGMC for MT mode? It is pretty straightforward for SD content and so worth it. My system chews through SD content in almost real time at the slow setting.

[CZbwoi]

But if I take a random 854x480 video, it pads to 856x480 afterwards if you notice the dimensions. Some filters require mod 8 or mod 16 dimensions, but usually not ones that work in RGB, usually only ones that work in 4:2:0 subsampled YUV . But I can replicate what you are seeing with a 854x480 input. A 848x480 (mod16) input doesn't exhibit that issue . So you might move the Spline36Resize(854,480) step later, after ccd , since you were starting with a mod16 video from vegas IIRC

Thanks, it's now the last step in my script and that fixed it, hopefully it'll come out alright.

ccd messes up with mod 2 frame sizes. Change it to 852x480 (mod 4) or 856x480 (mod 8) and the problem goes away.

Thanks, if PDR's fix doesn't work in the final export then I'll do this. I'll stick with 854x480 because that was the direct resize I ended up needing in this thread for these videos.

I hate to jump in and potentially confuse the OP, but why haven't you configured QTGMC for MT mode? It is pretty straightforward for SD content and so worth it. My system chews through SD content in almost real time at the slow setting.

If you look back through this thread (and maybe another) I tried it out and it ended up screwing me up n a couple ways. First it looked like it was going nice and fast, but then there were problems. Last I recall it wouldn't stop giving me this colored flash in my exports no matter what I did, so I switched back to the regular version.

[CZbwoi]

Sorry for the quick post back, but I just discovered another problem that comes out of using Avisynth (or in this case, AvsPmod). When I frameserve into it, it makes the picture overall darker and/or crushes blacks in some instances beyond belief. I took out every step besides the basics one by one and none of them changed the fact. I haven't noticed this til now when I decided to give a recent final video a watch all the way through.

What you see in Vegas, can clearly see the details in the suit:




Inserted into the script, everything in the suit is now one big black shape:




Now if I erase everything from ConvertToYV12(interlaced=true, matrix="pc.601") and on, the colors are back to normal and the suit isn't crushed...so I'm assuming it's something to do with this code being wrong. What's wrong with this part over here..? Anyone have any idea? I was only using this instance of the code because someone recommend it to me, but it looks like there's something it does wrong to my videos in the end.

[poisondeathray]

You can use Rec601

ConvertToYV12(interlaced=true, matrix="rec601")

But that shouldn't be right if you had 8 bit project in vegas with studio RGB sources like native camera clips. For those projects, PC.601 is correct . I don't know how you have vegas setup

[CZbwoi]

Woah, well that actually fixed it, you did it! Can you explain why this was the case? I mean they are native camera files and I never changed any of the Vegas defaults for things like this, I wouldn't even know where to look to be honest. But everything should be fine now I suppose from what it looks like.

[poisondeathray]

Did you have something like a levels studioRGB to computerRGB filter or vice versa applied anywhere in the project ?

[CZbwoi]

The only other things added to it other than what you see in the AvsPmod screen are what you see in the Vegas screen, White Balance and Color Curves (where you see it say Channel: RGB in it's setting..?) applied to the entire video track as 1 instance rather than to each individual clip.