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

[poisondeathray]

Stabilization always leaves black borders

You can either 1) leave the borders 2) zoom in 3) try to "fill" the borders by interpolating adjacent frames, or mirror blend edges 4) some combination of the above

You can adjust the amount of stabilization . For example, you might want to smooth it only a bit, so you don't have to zoom in so much; or you might choose to only smooth out really bad sections etc... A tiny amount of stabilization just to slightly reduce the hand jitters makes a huge difference in the average viewer's eye . It should be easy for you to test a small section, but I can post a comparison if you want




As for the PP interlaced 4:2:0 chroma mishandling, that's a pretty big FAIL in my book. I'll check later today to see if there is an easy fix. If you use a digital 4:2:2 intermediate it should work ok , but that's another step (...jumping through hoops again...)

[CZbwoi]

That's called "Jello Effect". If the footage is valuable, try to stabilize it by hand first to smooth out the extreme shakes. I don't speak the PP terminology, but in Vegas I pre-stabilize using the Pan/Crop tool, which can be keyframed. Surely PP has this tool, but under a different name.

A better way is to use "Motion Tracking" to stabilize. But I think you need AE to do that.

AE...Adobe Encoder?

But the good news is that after you deal with trying to stabilize via software, you will be motivated to keep a steady hand on the camera.

Lol, I do. This isn't my footage, it's my uncle's from when he went to Europe.

You can adjust the amount of stabilization . For example, you might want to smooth it only a bit, so you don't have to zoom in so much; or you might choose to only smooth out really bad sections etc... A tiny amount of stabilization just to slightly reduce the hand jitters makes a huge difference in the average viewer's eye . It should be easy for you to test a small section, but I can post a comparison if you want

I'll play around with it and see how it goes altering the settings.

As for the PP interlaced 4:2:0 chroma mishandling, that's a pretty big FAIL in my book. I'll check later today to see if there is an easy fix.

Do you think it has anything to do with how I'm exporting it? I mean I think you're pretty aware of all my export settings by this point, but you never know. For what it's worth, I also see it happening on the car freshener in the PP preview window, before the exports, frame by frame.

If you use a digital 4:2:2 intermediate it should work ok , but that's another step (...jumping through hoops again...)

Meaning export it again through another program or..?


Also if none of this chroma mishandling goes on when exporting the same way in Vegas I'll gladly switch, I just need to know if that's the case.

[LMotlow]

Yeah, got some chroma problems. This is a smaller copy of the image posted in #56:
Is the frame size really 1920x1081? Can't work with odd numbered pixels in YUV, bro'. I think that was mentioned.


This is the same frame from the attached mp4, at 16:9:


We could be here for a long time. Still some sawtooth edges to get to, looks like.

[But the good news is that after you deal with trying to stabilize via software, you will be motivated to keep a steady hand on the camera.

And don't shoot through glass.

[poisondeathray]

As for the PP interlaced 4:2:0 chroma mishandling, that's a pretty big FAIL in my book. I'll check later today to see if there is an easy fix.

Do you think it has anything to do with how I'm exporting it? I mean I think you're pretty aware of all my export settings by this point, but you never know. For what it's worth, I also see it happening on the car freshener in the PP preview window, before the exports, frame by frame.

No, it's there alright. It's a bug , even in the CC 2015 version (from your screenshots it looks like you're using CS6 ? )

If you use a digital 4:2:2 intermediate it should work ok , but that's another step (...jumping through hoops again...)

Meaning export it again through another program or..?


Also if none of this chroma mishandling goes on when exporting the same way in Vegas I'll gladly switch, I just need to know if that's the case.

Yes, I mean one of the workarounds, exporting something like a 4:2:2 lossless intermediate from another program. Because the bug exists in AME too (you can't use AME to batch transcode, because the exported files will have the problem)

Vegas works ok in terms of the 4:2:0 interlaced chroma mishandling

[CZbwoi]

Is the frame size really 1920x1081? Can't work with odd numbered pixels in YUV, bro'. I think that was mentioned.

No, VLC says that the display of the file is 864x482 and the display resolution is 854x480. That picture is from me going fullscreen in VLC and pressing the print-screen button.

This is the same frame from the attached mp4, at 16:9:

M2U00012_16x9DAR_480p.mp4

Yes, that actually looks better! Did you do this through Vegas? The only thing wrong here is that the borders are back, but I suspect that will be fixable doing the Square Pixels route I learned here, right? Should work the same as in PP...

We could be here for a long time. Still some sawtooth edges to get to, looks like.

Do you mean that sharpness we see everywhere compared to my recent frame, like around those 3 rectangle car grills in the bottom right? If so, what would you suggest?

No, it's there alright. It's a bug , even in the CC 2015 version (from your screenshots it looks like you're using CS6 ? )

PP CC 7.2.1

exporting something like a 4:2:2 lossless intermediate from another program

I still don't understand what this means. Take the file PP gave me and put it into another? "Link" PP with another program at the same time..?

Vegas works ok in terms of the 4:2:0 interlaced chroma mishandling

As long as LMotlow's export he provided is from using Vegas and we can fix the border problem on there too (and that sawtooth problem he was referring to) I'll switch ASAP, because I saw no red leaking in his file and that was great to see.

[poisondeathray]

"sawtooth edges" is a term describing deinterlacing artifacts. The look like stair stepping. They are aliasing artifacts. You get those when you deinterlace with NLE's , or handbrake/vidcoder

LMotlow is probably using avisynth, QTGMC, x264. Certainly vegas alone cannot produce clean deinterlacing like that - it only has 2 modes : blend and interpolate. Both are as bad as PP's deinterlacing. He altered the contrast a bit too. Notice the motion is "smoother" compared to the others? That's because it's double rate deinterlaced. The other things I would do personally if you had time is stabilize it a bit, fix the overbrights, maybe fix the exposure in the interior of the vehicle

exporting something like a 4:2:2 lossless intermediate from another program

I still don't understand what this means. Take the file PP gave me and put it into another? "Link" PP with another program at the same time..?

The problem with PP is upsampling the 4:2:0 interlaced source. So to get around that you upsample and encode it in another program to import into PP. 4:2:2 solves all the interlaced chroma handling problems in virtually every program that has those sorts of problems. Interlaced 4:2:0 can cause a lot of problems, because 1 chroma sample is spread between 2 lines - there are different ways to approach that and it can cause inconsistencies. I mentioned this earlier, but it looks like PP is upsampling the chroma as progressive, instead of interlaced. That's the cause of the artifacts you are seeing. EPIC FAIL for a "professional" program like that. Generic 4:2:0 interlaced MPEG2 might not be "pro" acquisition format, but it's certainly common enough that it should be handled properly

[budwzr]

Whew, ya'all smart fellers are gonna put a $10 cut on that?

All joking aside, that was a heck of a good job LMotlow.

[LMotlow]

"sawtooth edges" is a term describing deinterlacing artifacts. The look like stair stepping.

Your term is better. Referring to the metal panels at the windshield wipers. Kinda rough. But I don't think anybody would notice.

The other things I would do personally if you had time is stabilize it a bit, fix the overbrights, maybe fix the exposure in the interior of the vehicle

Yep, I tried all that. Just no time now. The interior is easy, but shooting through the windshield blew away contrast and color. If you light up the darks, stuff in the street gets more washed out. An exercise for masking in AE, I guess, or premiere or Vegas.

I got worn out with all the resizing and whatnot, but I guess one can do it if it suits. I used 720x480, QTGMC deinterlaced at "faster", then encoded 16:9 DAR. Added only SmoothLevels to darken blacks a bit and a touch of saturation with SmoothTweak. Could use a whole lot more than that though. I took the lazy way out.

[CZbwoi]

I mentioned this earlier, but it looks like PP is upsampling the chroma as progressive, instead of interlaced.

Well in PP's Basic Video Settings for the export I set the Field Order as Progressive...so is that a problem?

I did that because of what you said over here:

If you have the file and sequence interpreted as interlaced (ie as you do now), all you do is set "progressive" instead of field or TFF in the export options. Then it will deinterlace the export. If you have frame blending on it will blend deinterlace, if you have frame blending off it will interpolate. If you set framerate to 59.94p it will be double, 29.97 will be single rate.

That's how you told me to deinterlace the export and then double it by setting it to 59.94.

The problem with PP is upsampling the 4:2:0 interlaced source. So to get around that you upsample and encode it in another program to import into PP.

How would you recommend I do this? And I imagine I'd have to do this with every single original file (not the export I have now) and then load them all into PP?


LMotlow, so I take it you didn't do this in Vegas but in Avisynth? I would try to do it in there too but I don't know how I would manage loading ALL of my 12 videos into there at once, applying the filters to them at the same time and encoding them into 1 file when I'm done. That's my end goal.

[poisondeathray]

I would recommend you avoid PP for this one . Unless you're willing to do other steps. There is no combination of settings that "fixes" the chroma interpretation issue in PP alone with your native files. If you still wanted to use PP, you would batch encode to intermediate files, import those into PP.

You can load the clips into vegas, do you edits, levels/color/contrast manipulations as a project in vegas, frameserve out with debugmode frameserver, use avisynth. When you frameserve out of vegas, it is a "fake" AVI file that is 1 file. Avisynth "sees" that frameserved vegas project as 1 file. In avisynth you apply QTGMC, plus any other filters you want. This workflow is very solid and stable. Many vegas user use this dmfs+avisynth workflow, and there are step by step guides with screenshots on how to do it if you search. In this way you get all the benefits of using a NLE for editing , avoid the pitfalls of NLE deinterlacing , and all the benefits of avisynth processing (such as QTGMC), without large intermediate files.

One potential negative, is vegas doesn't come with a stabilizer. There are 3rd party ones that are addons in vegas you can use like mercalli, but not free. A good free one is virtualdub + deshaker, but it requires additional steps , intermediate files etc..., instead of being integrated like PP. (There are other minor things like vegas works in studio RGB, but don't worry about that). You weren't too keen on stabilizing initially , but I think if you did some tests with a small amount you will change your mind. So if you were going to do those other steps you're back to PP might be ok,

There are many other variations on how you could do it. Project planning is the most important step and will save you lots of headaches later on. For example, some people would batch stabilize the clips first, then import. But if you're going to be cutting out lots of footage, then it makes more sense to stablize later in the workflow, because you don't want to waste time on filtering etc... on footage that is just going to be cut out anyway

So yes, it's a few other steps too - whether you choose vegas or PP if you want to do it properly . Recall I said all NLE's have quirks and workarounds. You don't "have to" do those things. It's up to you what is "acceptable" to your eyes, what steps you're willing to cut out. Vegas's deinterlacing might be "good enough" for your goals. So run some small tests

[budwzr]

One potential negative, is vegas doesn't come with a stabilizer.

Yeah, it does. It's in the Media Effects buss. It's not very good though.

[poisondeathray]

One potential negative, is vegas doesn't come with a stabilizer.

Yeah, it does. It's in the Track Effects buss. It's not very good though.

Nice! I never even knew about it . You learn something new everyday

But maybe "not very good", is "good enough" in some scenarios ? Certainly worth a try.

[CZbwoi]

I would recommend you avoid PP for this one . Unless you're willing to do other steps. There is no combination of settings that "fixes" the chroma interpretation issue in PP alone with your native files. If you still wanted to use PP, you would batch encode to intermediate files, import those into PP.

You can load the clips into vegas, do you edits, levels/color/contrast manipulations as a project in vegas, frameserve out with debugmode frameserver, use avisynth. When you frameserve out of vegas, it is a "fake" AVI file that is 1 file. Avisynth "sees" that frameserved vegas project as 1 file. In avisynth you apply QTGMC, plus any other filters you want. This workflow is very solid and stable. Many vegas user use this dmfs+avisynth workflow, and there are step by step guides with screenshots on how to do it if you search. In this way you get all the benefits of using a NLE for editing , avoid the pitfalls of NLE deinterlacing , and all the benefits of avisynth processing (such as QTGMC), without large intermediate files.

One potential negative, is vegas doesn't come with a stabilizer. There are 3rd party ones that are addons in vegas you can use like mercalli, but not free. A good free one is virtualdub + deshaker, but it requires additional steps , intermediate files etc..., instead of being integrated like PP. (There are other minor things like vegas works in studio RGB, but don't worry about that). You weren't too keen on stabilizing initially , but I think if you did some tests with a small amount you will change your mind. So if you were going to do those other steps you're back to PP might be ok,

There are many other variations on how you could do it. Project planning is the most important step and will save you lots of headaches later on. For example, some people would batch stabilize the clips first, then import. But if you're going to be cutting out lots of footage, then it makes more sense to stablize later in the workflow, because you don't want to waste time on filtering etc... on footage that is just going to be cut out anyway

So yes, it's a few other steps too - whether you choose vegas or PP if you want to do it properly . Recall I said all NLE's have quirks and workarounds. You don't "have to" do those things. It's up to you what is "acceptable" to your eyes, what steps you're willing to cut out. Vegas's deinterlacing might be "good enough" for your goals. So run some small tests

Okay, I just might now. I see Vegas 13 is the most recent release, is that one fine or are there known versions with it and frameserving out with Debugmode FrameServer? Or should everything run smooth?

I just don't understand how this will all work though...so I create a timeline with the clips, then I frameserve it out into AvsP (because God knows I would have no idea what I'm doing with just a blank notepad file), after I'm done with doing everything in AvsP...what am I left with? I can't export anything through AvsP from what I recall, it'll just give me the .avs file. And I'm not supposed to export through Vegas because then I don't get all the benefits from Avisynth I did...so what exactly goes on after that? Does the .avs file get inserted into Vegas and you then export through there? It's all just a little hard to comprehend considering the timeline would be in Vegas that I shouldn't export through it (lest we run into the same problems with chroma and deinterlacing PP gave me), other video settings I want being stuck in Avisynth...I don't get how this will work lol it defies all logic to me.


@budwzr: Thanks for that info, if I ever get to use Vegas I'll try it out.

[budwzr]

I'm using VegasPro12. I think even Ver.11 has stabilizer too.

[poisondeathray]

I've used dmfs extensively for years with vegas 8 and 9. I only recently upgraded, so I don't know how stable dmfs is with the new version. I guess that's why I didn't know about the stabilizer either. I mentioned this earlier but PP has frameservers also, but they aren't as stable in newer versions

Typically one would use an encoder or GUI that accepts avs files. Some popular ones e.g. megui, ripbot, xvid4psp, staxrip, many others. Also many CLI encoders are typically compiled with avs support eg. x264, ffmpeg .

But avspmod can encode from the menu, tools => script encoder with CLI or VFW output , you just have to setup the paths and exe's in the options

[budwzr]

That's called "Jello Effect". If the footage is valuable, try to stabilize it by hand first to smooth out the extreme shakes. I don't speak the PP terminology, but in Vegas I pre-stabilize using the Pan/Crop tool, which can be keyframed. Surely PP has this tool, but under a different name.

A better way is to use "Motion Tracking" to stabilize. But I think you need AE to do that.

AE...Adobe Encoder?

After Effects.

Any method is going to eat up the edges. So you may want to upsample 2X first.

[budwzr]

I just got an email from Adobe asking for my input and opinion about their products, but I don't own any Adobe products. Hmmm...

[LMotlow]

I just got an email from Adobe asking for my input and opinion about their products, but I don't own any Adobe products. Hmmm...

Fill it out anyway, LOL!

The violent motion in the MPG can be smoothed a bit, but overall it ain't worth it IMO. Make it really smooth, you lose maybe 40% of the picture. DeShaker lets you do it without borders, which means resizing with artifacts, some smearing and whatnot. I put the earlier mp4 thru DeShaker at about 1/3 power and set for no borders. At least it smoothed most of the hard bumps from Honda Hatchback down to a 1976 Buick.

Can't brighten the back seat passenger very much without blowing away the street.

[smrpix]

Can't brighten the back seat passenger very much without blowing away the street.

You can if you add a soft mask, but....

Is it worth it? These are home movies and from my perspective, there's no reason not to allow them to look like home movies. I have nothing against color correction or a little fixing and cleaning and even some stabilization per se. But I also think there's a beauty and historical value in retaining what these particular cameras recorded in that particular way at that particular time. [/soapbox]

[budwzr]

...These are home movies and from my perspective, there's no reason not to allow them to look like home movies. I have nothing against color correction or a little fixing and cleaning and even some stabilization per se. But I also think there's a beauty and historical value in retaining what these particular cameras recorded in that particular way at that particular time. [/soapbox]

Yeah, it is what it is. A documentary.

[LMotlow]

All true. I made very mild corrections. It'll never look like a car chase from a Bourne movie.

[poisondeathray]

Yes, everyone has different goals. But I think stablized version is more pleasant to watch. Grandma won't ask "Sonny, why your hand shake so much?" . (I know it's the OP's uncle's camera, so just joking)

My opinion is you archive and watch the original if you want to see the original. This stuff is pretty basic for a NLE (basic levels and color correction, basic stabilization) - It's not that difficult. My opinion is if you're going to take the time to edit it, you might as well spend a bit more time fix it up a bit. You decide where you want to draw the line. Otherwise just append them as is as I suggested earlier, maybe trim a few bits out. Very fast, no quality loss. I have no idea why videoredo didn't work for you, it's usually very good mpeg2

One way to see if you like changes in a script is you can compare them side by side in a script and view with 32bit MPCHC directly, without having to encode. avspmod can compare different scirpts in different tabs, but you cannot PLAY it in realtime with avspmod

L-smash is nice because it doesn't require indexing for MP4/MOv, unlike ffms2. The OP is learning avisynth so I'm allowed to post a script.

a=LSmashVideoSource("LMotlow_M2U00012_16x9DAR_480p.mp4").spline36resize(854,480)
b=LSmashVideoSource("LMotlow_M2U00012_480p_DeShaker.mp4").spline36resize(854,480)
stackhorizontal(a,b)

Or sometimes is nice to see a half/half view

a=LSmashVideoSource("LMotlow_M2U00012_16x9DAR_480p.mp4").crop(0,0,-360,0,true)
b=LSmashVideoSource("LMotlow_M2U00012_480p_DeShaker.mp4").crop(360,0,0,0,true)
stackhorizontal(a,b).spline36resize(854,480)

For people that don't know avisynth I included encoded square pixel versions below of the comparison for LMotlow's videos, no audio. (Looks like some matrix , color differences, between the 2, maybe RGB in deshaker wasn't accounted for, but I realize it's just a demo). Only pay attention to the motion characteristics, because this is a lower bitrate multigenerational copy. Yes, it's pros/cons, and maybe for the OP's goals he doesn't want to zoom in even a bit, but based on those motion characteristics I would stabilize it a bit.

I'm not saying you should do this, I'm just throwing out some ideas . Obviously the passenger in the back is an important subject, at least in this clip, so personally I would try brighten it up a bit

To brighten the backseat, several ways you could do it - with curves, most NLE's have some sort of shadow adjustment (e.g. shadow/highlight filter), or even through masks/compositing but it's more tedious that way. More difficult to do in avisynth, because you have to Trim() or apply through a range and you can't easily "keyframe" values. But to do it in avisynth alone you could use masktools with a shadow mask to cover a Y range, and use mt_merge. Dogway has a nice lumamask() masktools wrapper function where you can pick the range, so you can selectively filter a range differently

Here is a rough example in avisynth to demonstrate , non PAR corrected (ie. 720x480 1:1). All I tried to do is compress the range to 16-235 and brighten the shadows, as you can see from the waveform (It's just histogram() in avisynth but flipped). It's easier to do levels, color manipulations in a NLE for most people, because there is a GUI and most importantly , easier to keyframe or adjust over time. Because these settings here wouldn't be valid in other parts of teh clip.

original


reduce highlights, brighten shadows



But pay attention to the blown out areas on the street (you can see pavement details now that were "not seen" before). Interior is a bit brighter, but you can only go so far before you reveal "dirt" and compression artifacts from the shadows, so you might have to denoise the shadows depending on how far you manipulate it, and you usually have to color correct it a bit afterwards too - I didn't do any of those, it's just a quick levels demonstration

The point is try not to discard details that are recorded. One benefit of vegas, is it uses studio RGB for most native camera formats, instead of "regular" sRGB. This means coming from a YUV source, you won't clip those overbrights right away when working in RGB. PP has a YUV capable timeline, so you can "fix" them as well. But some other editors convert to sRGB right away without allowing access to YUV and you lose those detailsdetails.

[LMotlow]

Good points, Poisondeathray. The quick way I used was hdragc but toned it down with heavy adjustments and SmoothLevels and then a touch of ColorMill later. Yeah, those street details would take time, they kept going in and out depending on what I used. Curves would be more precise. Usually I would've gone with a mild contrast mask and curves. Great filter.

As you suggested, there is low light noise in the MPG's shadows. EZDenoise took pretty good care of that. Thanks for the new demo.

[CZbwoi]

I've used dmfs extensively for years with vegas 8 and 9. I only recently upgraded, so I don't know how stable dmfs is with the new version. I guess that's why I didn't know about the stabilizer either. I mentioned this earlier but PP has frameservers also, but they aren't as stable in newer versions

So this is something I need clarification on: I import my files into Vegas, set it up on the timeline how I want, cut stuff out, apply some color filters, maybe stabilize, etc. Then I frameserve it out into AvsP, apply QTGMC and maybe one other thing, and then I'm left with only the choices of exporting it through there through x264 or VFW which gives me an .avi. Right? Because I can't really just save the script and insert it into something else since the "source file" is frameserving only into AvsPmod, there is no completed file, I simply inserted Vegas' timeline into AvsPmod so I have to complete and export it through there since there is no actual file to encode the .avs script with. I got this all correct right, or is there another way to export/encode a frameserved Vegas timeline after you're done with it in AvsPmod?

Also, is there a way to add audio muxing capabilities into AvsPmod when using their x264 encoder? Or some way to make it so that it adds the audio into the .mp4 on it's own when it's done with encoding? Or would I have to do that myself with Audacity, mp4box?

Speaking of, how would I get the audio from the timeline to mux with my newly made .mp4? I would have to export the project from Vegas just so I could rip it's audio to mux it?

Any method is going to eat up the edges. So you may want to upsample 2X first.

Can you explain what that means how to do this? And it wouldn't be necessary to upsample if I were to frameserve it to AvsPmod, correct? As it'll process and and encode it without those weird things NLE's do?

Is it worth it? These are home movies and from my perspective, there's no reason not to allow them to look like home movies. I have nothing against color correction or a little fixing and cleaning and even some stabilization per se. But I also think there's a beauty and historical value in retaining what these particular cameras recorded in that particular way at that particular time. [/soapbox]

That's how I felt sort of about the stabilizing bit, yeah it fixes it but ultimately it takes away from how it was filmed, the moment... Not to mention that a lot of the time you'd have to lose a significant amount of picture to achieve that effect and zoom in, and when it's not great quality to begin with...idk about that. But I'm still open to trying it out and experimenting, even if I would implement it a little bit to smooth it out. I appreciate all the examples and explanations you guys have posted of how to do it and what it would look like in action, thanks!

All true. I made very mild corrections. It'll never look like a car chase from a Bourne movie.

Lol. Speaking of, I'm hyped for the new movie so much.

Otherwise just append them as is as I suggested earlier, maybe trim a few bits out. Very fast, no quality loss.

Yeah, but then these videos wouldn't have the wondrous capabilities of QTGMC

L-smash is nice because it doesn't require indexing for MP4/MOv, unlike ffms2. The OP is learning avisynth so I'm allowed to post a script.

Are we...usually not allowed to post scripts here..? Like we can't post .avs scripts or..?

One way to see if you like changes in a script is you can compare them side by side in a script and view with 32bit MPCHC directly, without having to encode. avspmod can compare different scirpts in different tabs, but you cannot PLAY it in realtime with avspmod

So I'd just have to DL LSMASHSource and insert it into Avisynth's plugins, insert that code you posted somewhere inside MPC (can VLC do this?) and it will give me a live comparison with my script? That's awesome.

File Type: mp4 LMotlow_deshaker_compare.mp4 (55.99 MB, 4 views)
File Type: mp4 LMotlow_deshake_compare_halfview.mp4 (28.64 MB, 4 views)

That whole post and your examples were awesome. Thanks for taking the time to do that and explain it, I will probably do the same now!

As you suggested, there is low light noise in the MPG's shadows. EZDenoise took pretty good care of that. Thanks for the new demo.

Okay, what I don't really get is this: do you guys always insert QTGMC() as a line and then you also add stuff like EZDenoise because QTGMC's auto-denoising didn't quite do the job? Or do you just use EZ and not QTGMC? Like what exactly did you code in regarding denoising/those 2 filters?


Thank you to all of you guys for your input, samples and explanations!

[poisondeathray]

So this is something I need clarification on: I import my files into Vegas, set it up on the timeline how I want, cut stuff out, apply some color filters, maybe stabilize, etc. Then I frameserve it out into AvsP, apply QTGMC and maybe one other thing, and then I'm left with only the choices of exporting it through there through x264 or VFW which gives me an .avi. Right? Because I can't really just save the script and insert it into something else since the "source file" is frameserving only into AvsPmod, there is no completed file, I simply inserted Vegas' timeline into AvsPmod so I have to complete and export it through there since there is no actual file to encode the .avs script with. I got this all correct right, or is there another way to export/encode a frameserved Vegas timeline after you're done with it in AvsPmod?

Also, is there a way to add audio muxing capabilities into AvsPmod when using their x264 encoder? Or some way to make it so that it adds the audio into the .mp4 on it's own when it's done with encoding? Or would I have to do that myself with Audacity, mp4box?

Speaking of, how would I get the audio from the timeline to mux with my newly made .mp4? I would have to export the project from Vegas just so I could rip it's audio to mux it?

You have to think about what stabilizer you want to use if you're going that route (according to budwzr, vegas' internal one isn't good), and how to integrate it into your workflow. But yes, you can export with avspmod, or use another program to import the .avs from avspmod. Yes, Vegas has to be open with the project open all this time when frameserving (other no frames will be served) . Another option to get it out of vegas is to use a lossless intermediate (e.g. lagarith, ut video codec, magic yuv) . Pros/cons - when you do multipass encoding, or have heavy scripts / filtering, it actually becomes faster and more stable when using a lossless intermedate. The main drawback is you need lots of storage space.

Audio export from avspmod isn't great, maybe that's why people tend not to use the export from avspmod's menu. The way you would typically do it is frameserve both audio & video from vegas (the fake AVI signpost contains both if you checkmark audio). You load audio into the script (it's automatically done with AVISource) . You need to convert to YV12 with a PC matrix to "undo" the studio RGB that vegas uses (the levels will look correct), and specify the interlaced=true switch (otherwise you will get chroma artifacts similar to the ones in PP) . That .avs will contain both A/V so you can plug that into some GUI like one of the ones listed earlier and it will do everything for you, audio,video, mux. x264 is really a video encoder, so no audio itself. There are some custom builds that have audio, but not newer ones. ffmpeg can also encode audio, video and mux at the same time. If you wanted to , you could export audio directly from vegas and mux it in yourself, but probably easier to include it in the signpost and let some GUI do it for you , at least at first

AVISource("signpost.avi")
ConvertToYV12(interlaced=true, matrix="pc.601")
AssumeTFF() #your source was TFF, but it might be BFF for other sources
QTGMC(whatever settings)
#other filters here
.
.

Otherwise just append them as is as I suggested earlier, maybe trim a few bits out. Very fast, no quality loss.

Yeah, but then these videos wouldn't have the wondrous capabilities of QTGMC

Yes, but another option is to playback with QTGMC in realtime. And you should be archiving the originals. But it's more difficult for non techie family members to set up avisynth, scripts etc... so it's more friendly to produce a video that has everything "baked" in and you can just click it to play

L-smash is nice because it doesn't require indexing for MP4/MOv, unlike ffms2. The OP is learning avisynth so I'm allowed to post a script.

Are we...usually not allowed to post scripts here..? Like we can't post .avs scripts or..?

No, I'm just joking - it's just that some people don't like avisynth or scripts. I was scared of it and stayed away from it for years. But it's worth the time learning, very useful for many types of A/V manipulations.

One way to see if you like changes in a script is you can compare them side by side in a script and view with 32bit MPCHC directly, without having to encode. avspmod can compare different scirpts in different tabs, but you cannot PLAY it in realtime with avspmod

So I'd just have to DL LSMASHSource and insert it into Avisynth's plugins, insert that code you posted somewhere inside MPC (can VLC do this?) and it will give me a live comparison with my script? That's awesome.

You can open .avs scripts with MPCHC to play. You save your script, and to MPCHC it "looks" like a video. VLC typically can't do it, unless it's a special build with avs support. But realtime playback also depends on your hardware, the overhead in vegas (eg a bunch of vegas filters), what filters you use (QTGMC can be very slow, if you temporarily use a fast setting it might help) etc... The more "stuff" the less likely you will get realtime playback (too much CPU load and overhead, thread bottlenecks,e tc..). If you used a lossless intermediate, you're more likely to get realtime because all the vegas changes are "baked" into that intermediate. Also when you get more comfortable, look at avisynth MT, because you can get a nice speedup 2-3x with QTGMC alone. But it's more "dangerous" to use the more temporal filters you use, and if you add overhead like frameserve from vegas - there is a higher risk of mixing up frames and crashing, memory problems. You can read up on it when you have time.

Okay, what I don't really get is this: do you guys always insert QTGMC() as a line and then you also add stuff like EZDenoise because QTGMC's auto-denoising didn't quite do the job? Or do you just use EZ and not QTGMC? Like what exactly did you code in regarding denoising/those 2 filters?

It's one of those subjective things, personal taste. EZDenoize is an option but there are literally hundreds of other denoising filters and approaches. Sometimes one is better suited for some situations than others. That;s why you have avspmod to compare and MPCHC for realtime comparison. For example, you have a bit of chroma noise too (color noise). QTGMC on the slower settings reduces it a bit, but some people might use a chroma denoiser

My advice is try a few things out, but try not do overdo denoising. It will end up looking worse. Everyone is guilty of that when starting out.

[budwzr]

Increase the horizontal PPI by X2 allows you to do a Ken Burns. You'll be able to crop the image more after stabilize. Or you could say "increase the zoom factor".

The upsample size is inversely proportional to the crop amount. One gets bigger, the other gets smaller.

And run an Unsharp Mask on the upsample too.

[CZbwoi]

Another option to get it out of vegas is to use a lossless intermediate (e.g. lagarith, ut video codec, magic yuv) . Pros/cons - when you do multipass encoding, or have heavy scripts / filtering, it actually becomes faster and more stable when using a lossless intermedate. The main drawback is you need lots of storage space.

Are these options to frameserve it out into AvsPmod or to turn it into a lossless .avi file to insert into AvsPmod?

That .avs will contain both A/V so you can plug that into some GUI like one of the ones listed earlier and it will do everything for you, audio,video, mux.

But the export will be the same size as the combined amount of videos I imported into Vegas or larger..? Will it always be the same size as the source combined files unless you use like x264 and shrink it down?

matrix="pc.601"

So I always add this bit after the interlaced=true when I'm frameserving from Vegas (assuming I don't manually change RGB and color stuff on my own like a moron)?

It's one of those subjective things, personal taste. EZDenoize is an option but there are literally hundreds of other denoising filters and approaches. Sometimes one is better suited for some situations than others. That;s why you have avspmod to compare and MPCHC for realtime comparison. For example, you have a bit of chroma noise too (color noise). QTGMC on the slower settings reduces it a bit, but some people might use a chroma denoiser

I meant to ask do people normally insert a line that says just QTGMC() and then another line that uses EZDenoize from QTGMC? Does the regular insertion of QTGMC() as so have no denoising features so you're forced to add one that has EZDenoize? Or does QTGMC() have only a little amount of denoising done so you further do it again with EZ or another? And is it a good idea to insert two instances of it, one as the regular line with nothing added and one with EZ?


Thanks for all that info btw, I'll try it all out tonight!

Increase the horizontal PPI by X2 allows you to do a Ken Burns. You'll be able to crop the image more after stabilize. Or you could say "increase the zoom factor".

The upsample size is inversely proportional to the crop amount. One gets bigger, the other gets smaller.

And run an Unsharp Mask on the upsample too.

I'm assuming this is for Vegas, PP or other NLE's though, right? I wouldn't need to do this in Avisynth/AvsPmod unless I'm stabilizing correct? If I don't do any of that and frameserve it to AvsPmod I won't get sawtooth?

If that's not the case, how do I upsample in avisynth or Vegas? Is doing it in one better than the other?

[poisondeathray]

Another option to get it out of vegas is to use a lossless intermediate (e.g. lagarith, ut video codec, magic yuv) . Pros/cons - when you do multipass encoding, or have heavy scripts / filtering, it actually becomes faster and more stable when using a lossless intermedate. The main drawback is you need lots of storage space.

Are these options to frameserve it out into AvsPmod or to turn it into a lossless .avi file to insert into AvsPmod?

Those codecs are lossless codecs ; so NOT frameserving out with dmfs, but rather a physical intermediate

That .avs will contain both A/V so you can plug that into some GUI like one of the ones listed earlier and it will do everything for you, audio,video, mux.

But the export will be the same size as the combined amount of videos I imported into Vegas or larger..? Will it always be the same size as the source combined files unless you use like x264 and shrink it down?

If you use a lossless intermediate it will be several times larger. If you use dmfs, the "dummy" avi signpost is a few kb. The final export from x264 depends on the bitrate (filesize = bitrate x running time)

matrix="pc.601"

So I always add this bit after the interlaced=true when I'm frameserving from Vegas (assuming I don't manually change RGB and color stuff on my own like a moron)?

For projects with your native camera files, yes . Because vegas will be using a studio RGB conversion on those

For other input filetypes - maybe not, or you might have to use computer RGB to studio RGB effect on those instances

I meant to ask do people normally insert a line that says just QTGMC() and then another line that uses EZDenoize from QTGMC? Does the regular insertion of QTGMC() as so have no denoising features so you're forced to add one that has EZDenoize? Or does QTGMC() have only a little amount of denoising done so you further do it again with EZ or another? And is it a good idea to insert two instances of it, one as the regular line with nothing added and one with EZ?

If you're using EZ, you usually it together . 1 instance. IMO you can get better results, more fine tuning with other denoisers

how do I upsample in avisynth or Vegas? Is doing it in one better than the other?

You don't really need to upscale anything, it doesn't affect the proportion of black borders or edge loss that you're going to get through stabilization. But upscaling is better in avisynth with nnedi3_rpow2 than vegas or PP

[CZbwoi]

You don't really need to upscale anything, it doesn't affect the proportion of black borders or edge loss that you're going to get through stabilization. But upscaling is better in avisynth with nnedi3_rpow2 than vegas or PP

Upscaling = Upsampling?

And hold an a minute, are we only getting those "sawtooth edges" (seen in post #63) because we're trying to stabilize it? Assuming that's what you mean here by 'edge loss'.

And can you address this:

But the export will be the same size as the combined amount of videos I imported into Vegas or larger..? Will it always be the same size as the source combined files unless you use like x264 and shrink it down?

Like if you just export the .avs through a GUI without altering anything in there or inserting a codec like x264 will the size equal the combined size of the files you inserted into Vegas?

[poisondeathray]

You don't really need to upscale anything, it doesn't affect the proportion of black borders or edge loss that you're going to get through stabilization. But upscaling is better in avisynth with nnedi3_rpow2 than vegas or PP

Upscaling = Upsampling?

Upscaling is more generic term ; it usually refers to enlarging the frame dimensions.

"Upsampling" usually specifically refers to "chroma upsampling" , so in that case technically you can think of it as "upscaling" the chroma. You can read up about chroma subsampling, but the short explanation is your files are 4:2:0, that means the color information in the U, V planes is half width, half height - so 360x240 for a 720x480 frame. But the black/white information in the Y plane is full 720x480. Upsampling the chroma to 4:2:2 would mean full height, half width, so 360x480 of color information, the same 720x480 of black/white information

You were responding to some of budwzr's comments, and I just interjected saying you don't need to upscale anything for stabilization. It doesn't affect how much black border you have or proportionally how much you have to zoom in

And hold an a minute, are we only getting those "sawtooth edges" (seen in post #63) because we're trying to stabilize it? Assuming that's what you mean here by 'edge loss'.

I don't think LMotlow stabilized anything in Post 63 yet, so it's not "because". The edge loss from stabilization refers to the peripheral black borders around the frame that you use something like zoom to get rid of. But I suppose if the zoom algorithm used was poor (e.g. nearest neighbor) , that could cause blocky, aliasing, stair stepping too

And can you address this:

But the export will be the same size as the combined amount of videos I imported into Vegas or larger..? Will it always be the same size as the source combined files unless you use like x264 and shrink it down?

Like if you just export the .avs through a GUI without altering anything in there or inserting a codec like x264 will the size equal the combined size of the files you inserted into Vegas?

The avs is just a text file, a few kb. It's just a reference to the actual videos, and manipulations you've applied. The bitrate of the final encoded video determines the size of the final video. Filesize = bitrate * running time . So if the total audio, video bitrate and container overhead were the same, you would have the same size assuming you didn't cut anything out (because running time would be the same). If you have 1/2 the bitrate, it would be 1/2 the file size. If you don't compress it with something , it will be uncompressed video (many times larger than the original), because as soon as you use avisynth or dmfs, they frameserve uncompressed audio and video.