Never deinterlace for Youtube documentaries?
 

Hi there! I posted this comment on a thread awhile back

but hadn't heard anything back, so I wanted to follow up (hope this is okay) and see what your thoughts were:

I wonder if a lot of the reason that I'm getting so many conflicting ideologies while learning about capturing video is that my uses for my captured video might be much different than that of most others... 90% of the tapes I'm going to be capturing aren't likely to require significant restoration, and I don't plan to burn them to DVDs in order to watch them on a TV. Instead, the majority of the tapes that I'm trying to capture are going to be incorporated into mini-documentaries that I'm working on, like my YouTube show, for example. Or some clips from the tapes I capture might simply be uploaded straight to YouTube, which again would require some deinterlacing. And everything I do will be viewed on modern, progressive screens - like HDTVs, phones, and tablets.

Yes, I absolutely do plan to keep the original, unaltered AVI file of any of the more important tapes I capture, like personal family videos or "very important"/rare tapes in the archive for my YouTube show, because I understand the importance of keeping the original... but otherwise, it seems like for 90% or more of my purposes, I'm going to want deinterlaced video in an easy-to-manage file format like MP4 that I can easily share to YouTube or incorporate into other video projects. Is that not correct?

Anyway, sorry to throw all this into this PM, haha. I've just been doing my best over the past few months to learn the best process for capturing the videos that I have in a way that works well for my projects, and I'm surprised at how often I feel like I'm "this close" to having it figured out... before I end up in yet another new-to-me topic (like my foray yesterday into SAR vs DAR) and realize I still have so much to learn, haha.

This question was asked in a private message. Rather than hide our tech advice in private conversations, Site Staff will often answer PMs (from any site) here in the digitalFAQ.com forum, so that others may read and benefit from our expertise. Please continue the conversation here. Either login or join as a Free Member, and we can continue troubleshooting your video, photo or web related issue. Thanks for understanding our tech Q&A policies.

Okay. :)

You're probably mostly getting conflicting opinions because a lot of them are clueless. Everybody thinks themselves as expert, but Dunning-Kruger is mighty strong in the video capture realm.

Your uses are not that unusual. In fact, I'd suggest the days of DVD are behind us, and restoration has always been a smaller % of overall capturing. I've worked with a lot of indy filmmakers, doing just what you are.

Not as common, because Youtube is so aggressive with takedowns/strikes these days. But a lot of homebrew (most of it bad quality) is stuck on Archive.org these days (but that will end eventually).

That's the correct action to take. :congrats:

The archive/master is interlaced, the latter copy doesn't have to be. And, as you state, is now normal to deinterlace -- as well as upscale (but NOT with Topaz, you can do better, and for free).

Video is a large topic. Anybody that thinks they knows everything is somebody you should ignore. I think you're doing just fine. I saw this Youtube video a few days ago, it's really on-topic here. For years, I've consistently said newbies "don't know what they don't know", and why I detest people, especially internet trolls, that think admitting weakness (a lack of knowledge) is a bad thing.

If you repeatedly feel "this close", it means you're constantly learning. Congrats. That's how you should feel your entire life. Always learning, never knowing everything.

Youtube has been changing over the years, if not over the months, They use to have decent bandwidth for uploaded SD videos, acceptable de-interlacing..etc. Not anymore unfortunately, They treat anything less than 1080p as garbage, not because they suck at it but I believe they do it in purpose to save resources for modern high view rate videos, So depends on when you asked your questions, the answers may no longer relevant.

https://forum.selur.net/thread-597.html I like the guide that is at the bottom of that link. If you get overwhelmed by aspect ratios remember that you can just do 1440 by 1080 for YouTube and move on. If you want to look into why things are what they are then you can.

There is no better coverage of the aspect ratio and its origins than the coverage John Hess did in this well done video of his:
https://youtu.be/3CgrMsjGk7k

Whoa... so if Dunning-Kruger applies here, then I must in fact be unwittingly BRILLIANT at video capture! :laugh2:

But no, you make a good point there, and that does make an awful lot of sense. I'll have to keep that in mind in the future when I'm receiving (and vetting) suggestions on what I'm doing with all this. Thanks!

Ah, upscaling... that's one of the next things I'm trying to learn about. And I assume it will tie right in with the other topic I'm currently trying to learn about - "correcting" my aspect ratios. I've been assuming that I can do both of these things in Hybrid at the same time that I'm deinterlacing, and it seems pretty obvious that I'll be doing these things under the "crop/resize" tab... but there are several checkboxes and drop-downs within that tab that I'm not quite sure what to do with. Is there a good guide on the topic of adjusting aspect ratios and upscaling you could point me to? Or... perhaps a guide for each, if these topics are less interconnected than I realize?

I'm basically looking to make my clips non-anamorphic 4:3 (whether that means a square pixel format or changing the sample aspect ratio - pardon me if I'm not even using some of these terms correctly), and then, on occasion, upscaling my clips to 1440x1080p for YT upload.

Thanks! That was a good video. I was aware of the Dunning-Kruger effect, but it's been awhile, so that was an excellent refresher. Definitely one of those things I plan to teach my kids about when they get a little bigger...

This quote should be pinned at the top of the forum or something. I honestly might print it and stick it at the top of my computer monitor, haha. Thanks, man.

Fantastic! I'm actually printed this guide for some future late-night reading as I write this message. This guide is right up my alley - not only will it be useful for my little video capturing adventures, but it's a subject that has always genuinely interested me, and I've been planning to one day produce an episode of my show all about the differences between Terminator 2's various home video releases and their aspect ratios. So thank you very much for this link!

Awesome! I saved this to my watch list for the next time I have some free time to kill at work, haha. Really looking forward to watching this for the same reasons I mentioned above - thank you so much!

So far, I think you're doing fine.

For example, you have an obvious weakness in your capture hardware (lack of frame TBCC), but you're aware of it, and you've been monitoring it. I would highly discourage that for archival type use, but you're just grabbing clips for documentaries, and this is monitorable for short durations.

Well, not so much. AR needs to be addressed before upscaling. Some people chose to smush it all together, but it's easy to screw up something that way. I prefer to just take multiple encoding steps, or manually script line-by-line in Avisynth (and that's how you can reduce encodes).

I vastly prefer to write the script manually, then drop the .avs as the source file to Hybrid. I have Avisynth installed (both x86 and x64), and use AvsPmod (both x86 and x64), to visually preview the manual scripts. But with only resize and upscale, you can go blind.

selur has done a great job with Hybrid, but I completely dislike the order of operations (or at least the clarity of orders). There are questions. Manual scripting = zero question.

Uh ... huh? :huh1:

- So you have tapes (VHS, whatever), which is 4:3.
- You want to encode for Youtube.

Assuming NTSC numbers here, PAL differs.

This is fairly easy.
- You capture 720x480, which is obviously not 4:3. Don't worrya bout what that AR is, who cares.
- You crop 16 pixels total from left/right. That may be 8 from each side, 4/12, or even 16/0 or 0/16. Doesn't matter, 16 total. Due to how cropping algorithms works, it'll be even numbers here, not odd.

That gives you proper image AR after a resize.

Now, if you still have noise/black on the screen, such as head-switching noise at bottom, then you must crop 4x3. For example, if it takes 10 pixels of crop to remove the head-switching noise, you'll need to crop some from left/right as well.

- 4x3 = 4 wide x 3 high
- We're allowed to round.

So 10 ~= 9, and 9 / 3 = 3
So "4 wide" x 3 = 12

You'd crop 10 from bottom, and total of 12 from left/right (6/6, or whatever)

Now, if you're thinking "OMG, but I'll lose mah picture!!!"
Well, that's why many of us just mask with black. (Note: cropping 16 still required for 4x3 resize for viewing.) We'd mask 12 pixels (round up from 10, so total 6 top, 6 bottom). The way a VirtualDub mask with is "crop, center, fill black". So it re-centers.

The problem comes when stupid people say "OMG, but mah screen isn't full, I see black bars". They can't be happy either way, they are morons. Those are the people that stretch stuff, and are oblivious to AR. Batman becomes Fatman.

So either you crop 4x3, or you mask to save what you can.

Does this verbose help you more than a math equation? :)
Still questions?

Good dad. :congrats:

More and more, kids need to know about misinformation, and how to vet information and sources. Never trust, always verify. Especially avoid/ignore or be wary of people that think they know everything, but in reality know nothing. There's no shame in saying you only know X-Y-Z (few things), and not others. For some reason, in our internet troll society, admitting lack of knowledge is too often mocked. But the mockers are the idiots, not the person self-aware of knowledge gaps. They (mockers) never learn, you (kids) will.

:)

Hopefully the math equations make more sense after the verbose non-math explanation.

I have this great comic somewhere. Teacher shows kid math on board, asks for answer. Kid says "Who gives a f---?" Teacher look, says "Huh, you're right". Sometimes equations don't matter, knowing how to use the math is what truly matters. Equations just help you verify after the fact.

Wow, haven't seen that in a while. Yep, that should help too. Though I think it's more from the viewing POV, the historical film buff POV. Not necessarily too helpful for understanding the encoding, especially not the analog crop requirements.

Just wanting to make sure, so for cropping, is having the Y value be half of the X value not valid as well?

An example being you crop 14 on the left, 18 on the right. leading to the X side having a total crop of 32, which is an even number and below the max value of 40 for a 720x480 source.

Dividing that by half leads to 16, also an even number. Which is the total value the Y side would be, by cropping the top by 6 and the bottom by 10 for example. That's also below the max value of 24 for a 720x480 source.

This is a time where I hate X and Y, I don't want confusion.
You have top/bottom and left/right. Whichever is X or Y, I don't much care here.

If you have 720x480 NTSC, then crop 8 now. You start the 4x3 crop at 704x480 NTSC, not 720x.

So you gave a left/right crop number of 32, but on 720.
Remove 16, so you're cropping 16.

The left/right = the longer side = the "4" of 4x3. To get the multiplier, div 4. So 16/4 = 4
The top/bottom = 3 (of "4x3") * the multiple 4, so 12

Knowing head-switch noise, you'll probably crop all 12 from bottom, 0 from top.
And I'm guessing 8 on the sides beyond the 8 of 720>704?

After the cropping, you resize. Don't worry about x16 multiples, or anything like that. What matters is you arrive a 4x3 image ready to display at whatever size, probably at least 1280x720 (simple double 640x480, a 4x3) for Youtube.

The cropping 720>704 and resize-cropping is the hard part. Resizing numbers are not. The algorithms for a good resize are the hard part (ie, the popular Topaz is crapware, Avisynth is what you want for quality).

I see.. I just hope that having the top/bottom crop be exactly half of the left/right crop is valid but it seems like it isn't? That's what I've been doing pretty much since day one here.

Let me try and see if my math is correct.

I crop 4 from the top, and 8 from the bottom. Total crop is 12

4/3 is 1.3. 12 x 1.3 is 15.6, and you round that out to an even 16. Meaning you'd crop 8 from the left and right (or any value that equals to 16) but that's for a 704x480 source. For 720x480 you'd basically add 8 to that value on both sides, so 16 and 16 for example, meaning the total is 32.

So it looks like I was correct on my left/right cropping, and only slightly off for my top/bottom crop?

"Still questions?" Oh yeah. :laugh:

I'm not sure what I'm not getting here, but something isn't clicking for me. (But THANK YOU for all this information thus far - it's tremendously helpful!)

I don't think it's the math. The math isn't hard for me.

Maintaining a proper aspect ratio also makes complete sense to me - whatever my final resolution is from any sort of cropping, the horizontal and vertical resolution needs to remain in a 4:3 ratio. I totally get that. And so I completely understand why sometimes adding letterboxing is necessary - say for example if I want to crop some of the image from the bottom to remove VHS head-switching noise, but I don't want to crop any of the image from the sides... I can instead add the letterboxing to "fill in" the missing lines of vertical resolution as a result of cropping the lines of VHS head-switching noise.



So far, this all makes sense. What I'm missing is the part where I "de-stretch" my image horizontally from the captured resolution of 720x480 to a proper 4:3 resolution of 640x480... so that, when I open the video file on my PC for viewing (or drop the file into DaVinci Resolve... or share it with a friend... or watch it on my Jellyfin server), it displays as a proper 4:3 video.

Again, right now all the cropping and stuff makes sense to me (I think). But I'm still ending up with a video clip that is stretched horizontally when I view it. So my understanding (from what I read somewhere) is that the video is captured in a way that the pixels are horizontally stretched, right? And so I need to tell Hybrid to... "de-stretch" the pixels, in order to "correct" the resolution from 720x480 to 640x480?

I have a feeling I'm drastically misunderstanding or misinterpreting something here... I'm just not sure what. :laugh:



EDIT: It looks like you explained re-sizing to 4:3 right after you said "this is fairly easy." But somehow I didn't allow it to be fairly easy in my brain, haha. To me, it sounded like you were explaining how to cut some of the image off the sides...? So I may have mis-read something there.

720x480 is non-square 3:2 pixels used for storage. Why doesn't matter. It just is.
640x480 (or 720x540) is the 4:3 used for viewing.

Do you store your pants on a hanger, or in a dresser? Either way, odds are, your pants are folder in half (2:1). But do you wear you pants folded in half? No. Your pants are unfolded to 1:1, before you slide your butt into them. :laugh:

The one difference is NTSC video is closer to 704x486 in a 720x480 palette. So 6 pixels are whacked off top/bottom, and 16 total padded to the left/right. (Sometimes 8 per side, some 16 on one, or 12/4, or whatever. No rule. Just 16 total.) Your DVD player/TV knows this needs to be done, it's an invisible action to you on those viewers. But you're viewing with Youtube, so you need to know, manually take action to pre-crop. Does that make sense?

Samples needed.

I find Hybrid resize/crop confusing. Everything else is awesome about the software.

We humans tend to do that. Easy is "too easy", we have to make everything overly complicated. I do my best to demystify (and de-BS-ify) this stuff. It's not really that hard. But it does sometimes require "letting go", not trying to understand why, just how, and reaching needed outcomes. A lot of "why" is trivia, unless you're a software/hardware dev (which you're not, you're just a peon user like the rest of us).

As has been said, a capture at 720 x 480 should then have some combo of 16 pixels cropped off horizontally to get 704 x 480. It is best to mask off any video head switching noise along the bottom (or sometimes for other noise along the top as well) with a black border in it's place rather than crop it to less than 480 vertical rows. When using AVISynth, this is done with crop and addborders functions. Plenty of examples found via a forum search.

Interlaced 704 x 480 is suggested AVI format to keep for archival "master" before making any modifications. The lossless AVI files have no concept of an aspect ratio setting. However, you can use settings inside VLC player, for example, to tell it what the Aspect Ratio should be (i.e. 4:3) when rendering the pixels during playback.

Non-lossless, compressed formats such as DVD (i.e. MPEG2) or H264 (i.e. MPEG4) do have an aspect ratio setting (e.g. the SAR flag in H264), which is information stored in the video file which can be used by the player to adjust the rendering of the pixels stored in the file from the native Pixel Aspect Ratio stored in the file to a different width and/or height so that it will be seen at the expected width and height ratio. The aspect ratio setting does not modify the actual pixel data itself. It is merely a setting made available to players that are capable of stretching/squeezing pixels from the original dimensions stored in the video file to a new aspect ratio during rendering of the frames on playback.

Unfortunately, not all players are designed to support the aspect ratio setting. If you find this is the case for any of the players you are using (and certainly for youtube) then a physical pixel resizing has to be done as previously mentioned. An example is AVISynth Spline36Resize. This will modify the dimensions of the pixel data stored in the video file from say a captured 720/480 aspect ratio (i.e. 1.5 horizontal to 1 vertical) to a proper 640/480 (i.e. 1.333 horizontal to 1 vertical). Now the video will work for players that don't support aspect ratio setting, because the pixel dimensions were physically altered and stored to the new resized file. The aspect ratio setting could still be set to 1:1 after physically resizing and compressing to say H264, but may not be necessary. I find myself doing the physical resizing in AVISynth, due to frustration with the lack of universal acceptance for an aspect ratio setting. I'd rather not modify the video with a physical resize, but it is better than the alternative of it not being properly rendered at the correct aspect ratio by poor video software.

Some possibly useful posts from the archives include
https://www.digitalfaq.com/forum/vid...html#post74958
https://www.digitalfaq.com/forum/vid...html#post74292

Here is a trimmed AVI sample of the video I'm dealing with. As you may see, it's not only stored at 720x480, but it opens and plays at a stretched-out 720x480 (at least, that's how it's playing on my end when I open it with MPC-HC).

... Perhaps there's something wrong with my capture process?

My goal is to filter this in Hybrid so that it's deinterlaced, adjusted to a proper viewing aspect ratio of 640x480 (or in some cases 1440x1080 for YT), and converted to MP4 for a smaller size and more universal compatibility.

I think I have the deinterlacing part down and understand it pretty well... (you may note that this file, for some reason, is detected as "bottom field first" by Hybrid - when I force it to "top field first," the deinterlacing turns out great). It's the aspect ratio I'm struggling with. And then I think I'll be ready to move on and start working through my pile of tapes. :woot:

Here's my attempt at using Hybrid on it. All I did was cropping (18 left, 14 right, 6 top and 10 bottom) deinterlace using QGTMC and resize it to 1440x1080 using Spline36

Edit - Accidentally left on denoising using KNLMeansCL, degraining using TemporalDegrain and sharpening using Content Aware Sharpen.. oops

My god, it's beautiful. :clap: How do I hit the "Thanks" button multiple times? :laugh:

Any chance you would mind sharing screenshots of the three tabs within your Crop/Resize tab, so that I could see how you have it all set up and what options you're using? Because, as a wise man once said:

I've tried following instructions that seem pretty much exactly what you're explaining here (I feel like I understand everything that's been explained to me pretty well), but... I just don't seem to get the right results, so I'm clearly setting something up wrong, or choosing the wrong drop-down menu option or something. I wonder if I messed something up at some point in the past that's keeping me from getting the right results now.

Hey, that's okay - these are actually things I still need to learn about and experiment with, haha. If you have any comments on what you like about those particular settings or when you like to use them, I'd be happy to hear them. I have a fair understanding of noise and grain, and from my years of photo editing I feel like I'll have a decent eye for recognizing when sharpening will be helpful vs. harmful in video filtering... but I haven't used any of those filters on videos yet, so any comments on them (if you have the time, of course) would be gratefully noted. :salute:

Oh, also! Here's another thing I've been meaning to figure out - how did you convert the AVI to MP4 with the audio still intact? Because MP4 won't accept the PCM audio of the AVI file, of course, so the only way I've found to convert from AVI to MP4 within Hybrid is to select "Ignore" under Audio processing:
Attachment 18549

And that, of course, leaves me with a video with no audio... but at least Hybrid will convert the file instead of refusing to do so.

SO... my plan for my AVI-to-MP4 conversion process (thus far) has been to make a separate audio file extracted from the original AVI, converted to another audio format like MP3 that MP4 will be content with, so that I can select "Add" instead of "Ignore" under the Audio processing section, and add in the converted audio file to recombine into the newly-converted MP4. It seems extremely tedious, I know... but it's the best I've come up with thus far after asking about it on this forum. Do you have any advice on that part of the AVI-to-MP4 process?

Thanks again!

EDIT: Sorry, this should probably be a seperate topic, huh?

You can make a WAV in Vdub. Input that into Audacity and edit the uncompressed WAV then input the Wav into Hybrid and convert it to AAC of 160KBPS.

True.

While I was at it I redid the sample in Hybrid, this time disabling the Denoise, Degrain and Sharpen avisynth filters. Comparison between that and the file that had those filters enabled by accident can be found here.

I know I could've done some chroma noise reduction + chroma shift, but I intentionally didn't include them to keep the filtering process more simple, by using less filters overall.

https://imgsli.com/MzIzMzI4

Here's the crop and audio tabs. Now how I've always done cropping is keeping the top and bottom value equal to half of the right and left value. (10 + 6 = 16, 14 + 18 = 32. 32 / 2 = 16) But apparently that's wrong. What I should? be doing is multiplying the top and bottom (Y) value by 1.3, then take that result to make the crop for the right and left (X). Afterwards I'd then have to crop off 16 pixels

Example being just cropping 8 from the bottom and 4 from the top, 12 x 1.3 = 15.6 Round that off to 16. Since the cropping numbers need to even I'd do right by 8 and left by 8. But I need to crop off 16 more pixels, so I'd do 20 on the right, 12 on the left for this specific source, as there is more black on the left side of the frame compared to the right.

Final crop before resizing to 1440x1080 would be

Right - 20
Left - 12
Top - 4
Bottom - 8

not counting the 16 pixels I needed to chop off to keep the aspect ratio 4:3, it was 16 on the X, 12 on the Y

16 / 12 = 1.3

It seems like what I need to do is first crop off 16 pixels total (720 to 704), then do the cropping calculations from there..

Attachment 18551 Attachment 18552

Interesting! This really helps me get a better idea of what I can expect from these filters. Thank you!


Everything you said here makes perfect sense to me, including the math, and I very much appreciate the screenshots! I see now that I was correct in my suspicions that I have a setting messed up somewhere, because even following your screenshot exactly, I still end up with a horizontally-stretched video, as you can see in the preview below:
Attachment 18557

So my guess is that I have a setting messed up... elsewhere in the program?

OR, perhaps it's the version of Hybrid that I'm using...? I see that yours looks slightly different than mine, and has a few other options. Not only that, but I've noticed that more and more lately I'm having issues where clicking the little "shovel man" on the "Base" tab to export the newly filtered video file doesn't.... do... anything. So I wonder if this latest version has a bug somewhere that's getting in my way...?