Suggestions for video encoding improvements?

[CaseyStelken]

I captured a promotional Terminator 2 VHS tape from Japan and encoded it for sharing/viewing/streaming. I encoded a TON of different versions with different settings for file size/video quality/etc, and I finally settled on this MP4 that I'm feeling pretty good about!

I just thought before I moved on and started capturing/encoding other tapes in my pile, it would be nice to first find out if there's anything else I should consider... perhaps additional filters, etc?

I know every tape capture is going to require different filters, and this can all be fairly subjective... but I want to make sure I'm not overlooking anything obvious with this example before I move on, since I'm still pretty new at this and might not quite have a sharp eye for it yet.

Here is a snippet of the original capture file. As far as I know, it's fine... any concerns?
Terminator 2 Japanese store promo video [trimmed capture file].avi

Here is a snippet of the final delivery file: Cropped, resized to 1140x1080, and deinterlaced, with a bit of temporal smoothing and sharpening. Any suggestions?
Terminator 2 Japanese store promo video [trimmed delivery file].mp4

Here are the settings I used for encoding via Hybrid, if anyone is interested:
Capture settings 01.jpg

You must be logged in to view this content; either login or register for the forum. The attached screen shots, before/after images, photos and graphics are created/posted for the benefit of site members. And you are invited to join our digital media community.

Capture settings 02.jpg

Any feedback before I check this one off my list and move on is much appreciated! I've learned an awful lot here over the past few years, and I'm always ready to learn more - thanks, everybody!

[Aya_Rei]

I'd say additional filters are only needed if required to address other problems with the source. Piling on more and more filters can result in a mess that looks worse than the source material

Is that bitrate just for file size limits? Seems way to low in my opinion.

Bob deinterlacing was turned off so the footage was deinterlaced to 29.97 FPS, not 59.94 FPS. Granted since the promo video is mainly just 23.97 FPS clips from the movie, not having Bob on doesn't really matter in my opinion for this specific tape.

Perhaps it is best to keep the QTGMC settings at default (preset set to fast or faster instead of slow) and instead do sharping in Vapoursynth (using filters like LSFMod and/or CAS) maybe experiment with using those filters while the QTGMC sharpening is set to 1.0 and 0.1

I feel like filtering is like science, all about experimenting to see what works well.

[Selur]

• at least when I load it here, is does seem to be tff not bff (doesn't really matter when Bob is disabled)
• chroma seems to be shifted and/or bleeding
• right and bottom borders are discolored in some frames, moving BalanceBorders behind the crop and using might help
• personally, I don't use QTGMC for degrain or sharpening and usually stick with normal preset 'Faster', but with denoising and sharpening it really is a thing that comes down to personal preferences.
• I would not would let Hybrid calculate the resizing and not overwrite it. (if the input isn't properly flagged change the input par)

(slow and smooth: script)


Cu Selur

[CaseyStelken]

Quote:

Originally Posted by Aya_Rei

I'd say additional filters are only needed if required to address other problems with the source.

Ah, that makes sense!

Quote:

Originally Posted by Aya_Rei

Is that bitrate just for file size limits? Seems way to low in my opinion.

It is indeed for size (I have LOTS of tapes on my list that I'll be storing over time), but of course with quality heavily prioritized. This was after starting way up at 14,000kbps and gradually working my way down in bitrate to make sure the quality didn't suffer beyond the slightest amount.
I'm sure it'll vary from case to case depending on the video being encoded, but I was surprised to see a very, very minimal drop in quality or resolution even down to 3000kbps for this particular video. And believe me, I was overly nitpicky, comparing individual frames from different shots in the video for each new bitrate I experimented with. I've been messing with this video for a week or two now. Once I got below 3,000kbps was when I could see finally changes in quality that were noticeable enough that they weren't worth the smaller file size.

Is there any other reason I should consider a higher bitrate that I'm not aware of? The only other thing I can think of is if I were to upload this to YouTube... I know that YouTube does an awful lot of video compression - if I fed it a video with a higher bitrate, would I have a better chance at their compression not affecting the video quality as much as I would with this file, where the bitrate is right on the edge of compromising the quality?

Quote:

Originally Posted by Aya_Rei

Bob deinterlacing was turned off so the footage was deinterlaced to 29.97 FPS, not 59.94 FPS. Granted since the promo video is mainly just 23.97 FPS clips from the movie, not having Bob on doesn't really matter in my opinion for this specific tape.

My thoughts as well. From what I understand, Bob deinterlacing would only really be appropriate for content filmed specifically for TV broadcast or filmed via camcorder, right? News broadcasts, soap operas, home videos... that sort of thing, right?

Quote:

Originally Posted by Aya_Rei

Perhaps it is best to keep the QTGMC settings at default (preset set to fast or faster instead of slow) and instead do sharping in Vapoursynth (using filters like LSFMod and/or CAS) maybe experiment with using those filters while the QTGMC sharpening is set to 1.0 and 0.1

Good to know! I'll give those a shot. Any tips with using LSFMod or CAS? Or do you usually just use them with their default settings? I haven't used any of them yet... but I'll do some experimenting!

Quote:

Originally Posted by Aya_Rei

I feel like filtering is like science, all about experimenting to see what works well.

I hear you there, haha. That's why I've spent an excessive amount of time on this one video... just kinda figuring out how everything works and what sort of results I get from combining or adjusting different things. I figure I'll likely end up spending a lot of time on the next video I capture, too, and gradually I'll be able to spend less and less time encoding each video capture as I get to know what each clip needs, and as I gradually feel more confident that I've done the best I can with each project.

Thanks as always for the terrific feedback, man! I sure appreciate it.

[CaseyStelken]

Quote:

Originally Posted by Selur

• chroma seems to be shifted and/or bleeding

Oh really? Any tips on how I can recognize that? I thought the color looked okay, so this must be something I don't yet have an eye for.

Quote:

Originally Posted by Selur

• personally, I don't use QTGMC for degrain or sharpening and usually stick with normal preset 'Faster', but with denoising and sharpening it really is a thing that comes down to personal preferences.

Agreed on the personal preference part - however, my personal preference is certainly still developing as I'm quite new to this, so I appreciate your suggestions!

Quote:

Originally Posted by Selur

• I would not would let Hybrid calculate the resizing and not overwrite it. (if the input isn't properly flagged change the input par)

Looks like there's a typo at the beginning of your sentence here - would you mind clarifying? I want to make sure I pick up everything you're suggesting.

Quote:

Originally Posted by Selur

(slow and smooth: script)

I wish I could view your attached video file so I can learn from it, but the file won't open for me in any of my video players... I must be missing a codec? Not sure if this means anything to you:
MPC error.PNG

Also, I don't mean to sound like a total dope, but... where do people input handwritten scripts? I see people reference scripts all the time, but I can't find anywhere in Hybrid to enter them.

Thanks so much for all of the terrific feedback!

[CaseyStelken]

Also, I just realized that one of the screenshots I attached in my initial post above has some wrong settings shown. If it means anything to anyone, here were the actual settings I used for cropping/resizing/scaling in my encoded video:
Capture settings 02.jpg

[Selur]

video is av1 encoded works fine here. current mpc-hc plays it fine here too

Quote:

Any tips on how I can recognize that?

color halo on one side missing colors on the other usually means that the chroma is shifted.

Quote:

Looks like there's a typo at the beginning of your sentence here - would you mind clarifying? I want to make sure I pick up everything you're suggesting.

remove the first would, let Hybrid do the automatic calculation and feed it proper infos

[Aya_Rei]

Did my own restoration attempt, this time all in Avisynth, script is here

Image comparisons between all three https://imgsli.com/MzY2OTYx/0/1

Quote:

Originally Posted by CaseyStelken

Is there any other reason I should consider a higher bitrate that I'm not aware of? The only other thing I can think of is if I were to upload this to YouTube... I know that YouTube does an awful lot of video compression - if I fed it a video with a higher bitrate, would I have a better chance at their compression not affecting the video quality as much as I would with this file, where the bitrate is right on the edge of compromising the quality?

Good to know! I'll give those a shot. Any tips with using LSFMod or CAS? Or do you usually just use them with their default settings? I haven't used any of them yet... but I'll do some experimenting!

For bitrate, I keep it at 26000 kps for YouTube (since I resize to 1920x1440 so the videos get the vp09 codec), you can view the spec sheet here

For sharpening, I myself keep CAS at 0.60 as oppose to the default of 1, for LSFMod I take a page out of Lollo's book and just use the default "slow" preset.

[Feedbucket]

IMO: I'd say the chroma shift is the only thing that really needs help.

I noticed that your captures seem to have a similar noise profile with spikes in your levels - is this a GV-USB2 thing? Only people who spend time looking at histograms are going to be bothered by this though.

In general I'd caution against too much denoising (we probably both know what Cameron's opinion is). VHS has limited detail to begin with and removing an indication of absence of info also removes the implication that info could exist. If your goal is to accurately represent what you're capturing, consider how post processing might affect this.

My experience, x264 also does an amount of smoothing, so something else to consider if that's your delivery format.

Interesting that the blacks around the Japanese text are at Y=0. I'd leave the levels alone since the film behind it sits in range fine. EDIT: The blacks could maybe be bumped up a bit, but not on behalf of the text. output_lo=6 or so.

[CaseyStelken]

Quote:

Originally Posted by Selur

color halo on one side missing colors on the other usually means that the chroma is shifted.

Ah, now I see! Thanks so much - this is a major learning moment for me. I do recall learning in the past how VHS carries separate signals for luminance and chrominance, but I have never recognized chroma shift before, so I didn't know what to look for. Now that you've pointed it out, I'm seeing it all throughout my capture! So glad I learned about this now and not later - thanks so much!

Any tips on treating it? Under Hybrid's Filtering -> Vapoursynth -> Color -> Misc tab, I see "FixChromaBleeding," "ChromaShift," and "ChromaShiftSP." I tried "ChromaShift" and set "Right" to "2" and seem to be getting pretty good results. @Aya_Rei, is that what you used in your example? It looks similar to what I'm getting with this setting, so I think I'm on the right track.

Thank you both for getting me down the right path! I'm feeling pretty justified in obsessing over this video before moving on to others, haha.

Quote:

Originally Posted by Selur

I would not would let Hybrid calculate the resizing and not overwrite it. (if the input isn't properly flagged change the input par)

Quote:

Originally Posted by Selur

remove the first would, let Hybrid do the automatic calculation and feed it proper infos

So you're talking about the "Picture Resize" section under the "Crop/Resize" tab, right? The reason I checked the "Resize" box and set the dimensions to 1440x1080 is because I've been told that feeding YouTube a 1440p video would result in a better upload, because YT uses a better encoder for higher-resolution uploads...? [EDIT - I see that I misunderstood 1440p - see next paragraph.]

Quote:

Originally Posted by Aya_Rei

For bitrate, I keep it at 26000 kps for YouTube (since I resize to 1920x1440 so the videos get the vp09 codec), you can view the spec sheet here

26000?? That seems massive to me. I started at only 19000, which I thought looked about as good as it was going to look, and worked my way down until I saw a saw a discernable drop in quality. Then again, you are resizing to a bit larger of a resolution than I am.
...And as I type that, I'm realizing that perhaps I'm going for the wrong size. In my notes, I have the quote “If you're going to upscale 480i, you might as well go to 1440p, since it's 3X the resolution, so you get clean upscaling, no half-pixels or quarter-pixels.” (Not sure who said that, but it was elsewhere in this forum). And I'm realizing that I misunderstood 1440p to be 1440x1080... when it should be 1920x1440.

Now I'm really glad I made this post before continuing with my next tape...

[CaseyStelken]

Quote:

Originally Posted by Feedbucket

I noticed that your captures seem to have a similar noise profile with spikes in your levels - is this a GV-USB2 thing? Only people who spend time looking at histograms are going to be bothered by this though.

Ah, I bet I know exactly what was causing that. I probably had my router turned on when I made this capture, because I think I made this capture before I solved this issue I was having (only audible with certain tapes that had tracking issues... but apparently the spikes in levels were visible on the histogram on all captures?)

Quote:

Originally Posted by Feedbucket

In general I'd caution against too much denoising (we probably both know what Cameron's opinion is).

You're definitely speaking my language. I'm not trying to do any excessive clean-up or enhancement, and I'm not looking for a super-smooth image or to enhance what isn't already here... I'm just looking to capture and reliably present (in a progressive video format) a fairly faithful representation of what these tapes are supposed to look like when played back properly. I'm just trying to present what's already here on the tape as well as possible. I'm okay with maybe some minimal clean-up; a tiny bit of smoothing or sharpening here and there is okay, but I prefer that to be minimal.

Quote:

Originally Posted by Feedbucket

If your goal is to accurately represent what you're capturing, consider how post processing might affect this.

Absolutely. In other words, it's a VHS tape, and I don't mind it looking like a VHS tape.

[CaseyStelken]

Quote:

Originally Posted by Aya_Rei

Did my own restoration attempt, this time all in Avisynth, script is here

Image comparisons between all three https://imgsli.com/MzY2OTYx/0/1

Thanks so much for that comparison shot - that's super helpful for me to train my eyes for what other sorts of adjustments I can consider when working on these!

[Aya_Rei]

Quote:

Originally Posted by CaseyStelken

I have never recognized chroma shift before, so I didn't know what to look for. Now that you've pointed it out, I'm seeing it all throughout my capture! So glad I learned about this now and not later - thanks so much!

Any tips on treating it? Under Hybrid's Filtering -> Vapoursynth -> Color -> Misc tab, I see "FixChromaBleeding," "ChromaShift," and "ChromaShiftSP." I tried "ChromaShift" and set "Right" to "2" and seem to be getting pretty good results. @Aya_Rei, is that what you used in your example? It looks similar to what I'm getting with this setting, so I think I'm on the right track.

Thank you both for getting me down the right path! I'm feeling pretty justified in obsessing over this video before moving on to others, haha.

26000?? That seems massive to me. I started at only 19000, which I thought looked about as good as it was going to look, and worked my way down until I saw a saw a discernable drop in quality. Then again, you are resizing to a bit larger of a resolution than I am.
...And as I type that, I'm realizing that perhaps I'm going for the wrong size. In my notes, I have the quote “If you're going to upscale 480i, you might as well go to 1440p, since it's 3X the resolution, so you get clean upscaling, no half-pixels or quarter-pixels.” (Not sure who said that, but it was elsewhere in this forum). And I'm realizing that I misunderstood 1440p to be 1440x1080... when it should be 1920x1440.

Yes that's what I used, granted for Vapoursynth chroma shift shifts both chroma values, U (Yellow/Blues) and V (Reds) by your desired amount unless you specifiy it to do only one type (That was added in later Hybrid versions, needed to backport the script and manually add it in to my filter list). Though with ChromaShiftSP, positive numbers shifts it to the left while negative numbers shifts it to the right.

As for the bitrate, that high of bitrate, and upscaling to 1920x1440 is just to please YouTube. Judging from your notes you probably saw a post where fellow YouTuber Vwestlife recommends doing that.

For offline use I just do 1440x1080 with a bit rate of 15000 kbps. Granted it's probably best to not upscale nor deinterlace unless it it required, such as the case with YouTube.

[Selur]

As a side note, for chroma bleeding, using:
Filtering->Vapoursynth->Other->VSGAN or
Filtering->Vapoursynth->Other->vsMLRT
with 1x_BleedOut-Compact usually helps a lot.

Cu Selur