How/why capture 60 frames per second?
 

I mainly capture footage from VHS TV recordings from the early 2000s to the 80s. I'm 99.8% sure that the standard is 29.97 frames per second (or 59.94 fields per second) in NTSC. It's how I remembered watching TV and analog tapes growing up and unless there's a remaster of the actual film reels, there's absolutely no way to turn that into 60fps.

But, judging from what I'm seeing on YouTube uploads (some posted by those in this forum) it's sometimes 1080p/60 FRAMES and has an unnatural soap opera look. This is the .2% that I'm not sure what's happening.

Can someone solve this mystery for me?

"Capturing" 60fps is low-quality junk devices (and low-knowledge users).

Processing 59.94 fields into 59.94 frames, post-capture, is done to retain all motion in a full proper deinterlace. Noting that some footage fights field separation, lots of artifacts are the outcome.

1080p is a Youtube requirement, otherwise quality is further punished.

So "1080p59.94" is common.

The "soap opera effect" is so named because that's the way live and videotapes TV actually looked back in the analog era -- smooth, fluid video at either 50 fields per second for PAL or 60 for NTSC (later tweaked to 59.94 when color was introduced).

Yes, it takes two fields to form each complete frame of video, but those two fields do not represent the same moment in time, so with the proper de-interlacing methods, it is possible to preserve the full motion of the video and turn it from 50i or 59.94i into 50p or 59.94p.

Now, it is true that as far back as the 1950s with "I Love Lucy", a lot of TV shows were actually shot on film, not videotape, so they run at 24, 25, or 30 fps, and no de-interlacing method will be able to squeeze anything higher out of them. This extended into the digital era when a lot of shows began recording video at the "film look" frame rate of 24 fps -- ironically now including soap operas!

So, yes, there is a lot of TV programming on which it is impossible to produce true 50-60 fps out of them, aside from the "TruMotion" frame interpolation nonsense built into modern TVs and available in some video editing tools. But this should not confuse the fact that there are many decades' worth of video content on which that "too smooth, soap opera-like" 50-60 fps is exactly the way they looked when originally recorded and viewed.

Resolutions above 1080p are actually now the requirement for all videos and channels to get immediate access to the highest-quality encoding and full availability of lower resolutions. If you upload at 1080p or less, only "big channels" get those perks, and only if the video crosses a certain threshold of views (which is an unknown and moving target).

Thus I now upscale standard-definition NTSC video to 1440p, which is a nice clean 3X upscale from 480i. (1920x1440 for 4:3, or 2560x1440 for 16:9.)

Youtube will also label 59.94FPS or re-encode 60FPS, so just because you see the 60FPS, doesn't mean that the encoding or deinterlacing wasn't done correctly, just to turn each field into a frame.

The best video codec currently for youtube is supposedly VP9 and that doesn't necessarily trigger for smaller than 1440p uploads. For some reason, doing 1536x1152 DOES always result in Youtube encoding to VP9 (when displaying at 1080p) so that's what I use myself to keep the uploads a bit smaller, but still get VP9.

See attached screenshots where you can see under "stat for nerds" to see what the video you are watching is encoded at. I just used VWestLife's video he uploaded recently to show that if viewing at 1080p on his video, the 1080p copy is AV01 codec whereas mine with the odd resolution of 1536x1152 gets to use VP9 when displayed at 1080p. The actual file uploaded was x264 (in my case) @59.94FPS and then youtube decides what to encode after that. Also of note, if you go to 1440p or higher, then it'll kick into VP9, but I think 1440p is only available to youtube premium members and proably even extra overkill for original SD content.

Stats for nerds is also interesting in that you can see audio levels. Youtube will quiet down a video that is too loud, but won't increase the gain on a video that is too quiet. I've since started normalizing the audio to -1dB or so which gives max allowable volume at the loudest moment of the video.

Attachment 20539

Attachment 20540

This is what I observed as well. I don't consider myself close-minded so if there's a method to do 60 frames and not have it look unnatural and weird, I'm all for it. I've tried frame interpolation such as Flowframes and didn't like it. I tried deinterlacing through Yadif + bob + 59.94 frames on Handbrake and the results are often bad (see the Jockey commercial I attached).

So what you're saying is that while upscaling 480p to 1080p comes at the cost of making the footage blurrier, at least that quality downgrade is in our control and not leaving it in the hands of YouTube's software?

That's fair.

Maybe this depends by country then? Because in our country, everything moved in 30fps or less when viewed on a TV. Same goes for the VHS and Betamax tapes we rent.

Ah, this is what applies to our case then!

Is there a trick to not make your footage blurry when upscaling to what's essentially 3x the resolution?

I am familiar with the "too realistic/smooth effect" back when HDTVs first came out with increased frame rate features which I believe was due to poorly implemented frame interpolation.

The thing with 480i is that the there really is motion information in all of the ~59.94 fields as they are each from a different moment in time - that is unless the content had a lower starting frame rate like those shot on film. I don't at all notice a "false" look with my 59.94 deinterlaced captures.

This is just a random example from my channel that was deinterlaced to 59.94P with QTGMC. Does it look soap-opera-esque in your opinion?
https://www.youtube.com/watch?v=vtj3V6UNwi8

As far as your Jockey capture, it appears that you've got the field order reversed and/or there's a problem with the inverse telecine (as that commercial was likely shot on film and then adapted to NTSC later). You can see if you advance by single frames that many frames do not appear to progress in the right order.

That's the type of "60fps" I'm seeking! Unfortunately, it seems like this is impossible to replicate unless the source really does have that smooth motion. :depressed:

This is likely the case, as when viewed on the monitor directly from the VCR, the motion is 30 frames. Unless there's a similar copy out there that shows it to have been 60 frames.

What's your full capture chain and processing method usually?

Just for a more apples to apples comparison, this is one of my few "definitely older" set of commercials from an off air recording done in 1978.

https://www.youtube.com/watch?v=AxGoJVyouK4

While not exactly perfect on the motion blur from frame to frame, the sources for these were all ~24FPS film (being that old) as I suspect your Jockey example post was also. You actually don't want to use QTGMC for that, you want the inverse telecine setting to get it down to the ~24FPS that it started as. You can always "right click" on a video on youtube and see the framerate in "stats for nerds." So even if you have a non 60p stream, it could still be less than 30fps also and those might be the videos where you don't see as many issues.

Here's an example of what the "stats for nerds" looks like on that video - can see it's at 24FPS:
Attachment 20549


Also of note, you can have 60FPS video that doesn't actually have motion on all frames:
https://www.youtube.com/watch?v=IfV8M80LaDg

I think what happened there on that laserdisc was that the scaling DVD player was doing "live inverse telecining" along with the upscale which produces duplicate progressive frames every 2-3 frames. You can see if there's motion frame to frame in youtube by pausing the video and using "<" and ">" to advance or reverse by single frames.

I would have to guess that videos that do look like that probably weren't processed correctly or may not have been captured in 480i.

Sorry to go off-topic, but did you use your BVU-950 with a BKU-901 (internal TBC) to capture that John Deere ¾ U-matic tape?

Yes, that one was. The internal TBC is really unmatched in terms of dropout compensation at the downside of processing things in composite (and only able to output composite) which gives a smoother appearance which may or may not be preferred. You tend to get more "grain" or preservation of image noise which can give the appearance of there being more detail present with either Dub or modified S-Video output. Dub I'd say has more of that grain and modified S-Video is more of a middle ground in that aspect. I've never been impressed with external TBCs that have dropout compensation, but could be that I just haven't set the thresholds up correctly?

Ok, my question not being a huge YouTuber is why you want to force VP9 over AV1? I'm curious what the extra resizing step will do for me to force VP9?

This is kind of the answer I've seen:

AV1: The clear winner for visual fidelity and bandwidth savings. It is designed specifically for 4K, 8K, and HDR video. It utilizes advanced prediction and filtering, giving it a 30% to 50% reduction in file size over VP9 for the same visual quality.

VP9: Highly efficient compared to older codecs like H.264, but outpaced by AV1. However, in real-world streaming like on YouTube, heavily optimized VP9 files sometimes look sharper to the naked eye than default AV1 streams because YouTube encodes AV1 very quickly at lower bitrates to save server processing power.


So sounds like AV1 could be superior - IF - you could control the settings during the encode, which youtube doesn't let you do, or perhaps prioritizes bigger channels to get better encoding (which not gonna lie, makes sense to not spend a lot of encoding time on a video that is never likely to get more than 1000 views).

Also for quite a while it was normal practice to shoot on film and then transfer to videotape for editing and distribution, which will have you banging your head against the wall if you try to do inverse telecine on it, because almost every scene change in the resulting video will break the 3:2 pulldown cadence and ruin the whole thing, unless you want to chop it up scene-by-scene and basically redo all of the editing.

Such was the case with this video -- shot on what appears to be 16 mm film, transferred to and edited on videotape, and uploaded as "60 fps" (59.94) only because there's no easy way to convert it back down to its true original 24 fps frame rate:

Robin Williams visits Florence (1987 VHS)

Telecine... a problem so easily fixed if the original film masters would just be rescanned/reassembled and re-synchronized to the audio (unfortunately an expensive proposition, if the elements even survive).
TIVTC, VIVTC, and QTGMC in bob-mode + sRestore usually do a pretty good job with some sources, but other times it requires lots more manual tinkering, all of which I do in Hybrid nowadays.

Off topic but blended frames are the worst, it's the bane of most anime content.