Is 4k Any Better?

[DBounce]

A couple days ago I decided to do some test with the Canon C200 and Panasonic GH5S. I set up the two cameras next to one another and ran the C200 at 200 ISO vs the base 800 ISO of the GH5S. I then progressively increased the ISO of the Canon until it matched the GH5S, at which point I matched the same setting on both cameras. ISO 200 is the cleanest on the Canon. Both are quite clean at 6400 ISO. And this is good, as I almost never shoot higher than that. 

The real surprise came when I started to edit the footage in DaVinci Resolve 15 Beta. I'm using the free version as I see little point in paying for the full version since it comes bundled with the BMP4k. Well, as luck would have it, the free version has limitations. It cannot output at DCI 4k. The max output resolution is UHD. These days I tend to prefer Cinema Scope as it is wider and works well on high-end mobile devices with widescreen aspect ratios. With this in mind I selected DCI 2k Scope 2.39. Strangely, when I rendered out the footage... to my eyes it looked every bit as detailed as the 4k render from the GH5S. I checked to make certain I had rendered the Panasonic footage correctly. Sure enough, it was 4k. I am now left wondering if there is any reason to deliver anything in 4k given that DCI 2k looks so nice when down sampled.

Have any of you here experienced anything similar?

[EthanAlexander]

Bitrate being equal, yes they are very hard to spot differences in, but if you're delivering to web the platform will only stream at the bitrate they've chosen for each resolution. For instance, YT plays UHD at about 35Mbps and 1080 at 8Mbps, so you'd definitely see the difference there. 

[DBounce]

1 minute ago, EthanAlexander said:

Bitrate being equal, yes they are very hard to spot differences in, but if you're delivering to web the platform will only stream at the bitrate they've chosen for each resolution. For instance, YT plays UHD at about 35Mbps and 1080 at 8Mbps, so you'd definitely see the difference there. 

I may need to do some testing with YouTube then.

[webrunner5]

The paid version of Resolve 15 has a great noise reduction feature in it. But yeah if you are going to buy the 4K BMPCC wait, don't double buy Resolve.

And I think Ethan is right, just the data rate alone makes it worth going 2K or more on YT. 1080 at 8Mbps can be pretty Gnarly at times. But I have seen some damn good 1080p on there, but it might have been down sampled from 2K, 4K.

[andrgl]

If it's going to live on YouTube my belief is that you should be adapting to and using everything Google gives you (cards, subtitles, etc,) as the ranking algorithm will give you an edge. So yeah, I would upload to 4K and beyond if possible.

[Mark Romero 2]

9 hours ago, DBounce said:

A couple days ago I decided to do some test with the Canon C200 and Panasonic GH5S. I set up the two cameras next to one another and ran the C200 at 200 ISO vs the base 800 ISO of the GH5S. I then progressively increased the ISO of the Canon until it matched the GH5S, at which point I matched the same setting on both cameras. ISO 200 is the cleanest on the Canon. Both are quite clean at 6400 ISO. And this is good, as I almost never shoot higher than that. 

The real surprise came when I started to edit the footage in DaVinci Resolve 15 Beta. I'm using the free version as I see little point in paying for the full version since it comes bundled with the BMP4k. Well, as luck would have it, the free version has limitations. It cannot output at DCI 4k. The max output resolution is UHD. These days I tend to prefer Cinema Scope as it is wider and works well on high-end mobile devices with widescreen aspect ratios. With this in mind I selected DCI 2k Scope 2.39. Strangely, when I rendered out the footage... to my eyes it looked every bit as detailed as the 4k render from the GH5S. I checked to make certain I had rendered the Panasonic footage correctly. Sure enough, it was 4k. I am now left wondering if there is any reason to deliver anything in 4k given that DCI 2k looks so nice when down sampled.

Have any of you here experienced anything similar?

Just so I understand, both the footage from the C200 and the footage from the GH5s were rendered out as DCI 2K Scope 2.39 right???

And both were shot 4K DCI, right?

Or am I missing something???

[DBounce]

14 minutes ago, Mark Romero 2 said:

Just so I understand, both the footage from the C200 and the footage from the GH5s were rendered out as DCI 2K Scope 2.39 right???

And both were shot 4K DCI, right?

Or am I missing something???

I rendered the footage from the GH5S in 4k DCI, and the footage from the Canon C200 in 2k DCI. To my eyes the difference in detail and resolution was negligible. I also compared to previously rendered images from the C200 which were rendered in 4k DCI. Again little to no difference when comparing the 2k DCI to the 4k DCI. I'll go as far as to say I preferred the 2k DCI imagery. 

[webrunner5]

Are you looking at these on a TV or a 4K Monitor??

[kye]

Early last year I decided to "go 4K" buying a new MBP laptop, Dell 32" UHD display, XC10, and later on also upgraded my iPhone 6 Plus to iPhone 8.  (happy Christmas to me!)

I chose a 32" 16:9 display because after reviewing the THX and SMPTE standards for cinema displays (I think it was SMPTE, there were two standards anyway) they suggested that from my viewing position it was the right viewing angle.  I contemplated a 42" display, which would have been near the outside range of the two specs, and I also contemplated one of those 21:9 displays but they aren't full UHD so you can't see the full resolution on them.

What I realised was that the difference between 4K and 2K (or UHD and 1080) is that the higher resolution has 2x2 pixels for every single pixel on the lower res.  This may sound obvious, but think about it this way.

Unless you can see the individual pixels on your 2K / 1080 display from the viewing position, you won't be able to see any more resolution or detail in 4K / UHD.

Of course, as others have said 4K gets less YT compression (which I personally verified here if you're curious) but it also has some interesting benefits as a capture format even when you are exporting to a lower resolution.  One of these benefits is that in theory it turns an 8-bit colour space into up to an 10-bit colour space, but this is a complicated concept depending on many things, so it's probably better to think of it as potentially being better but not necessarily.

[webrunner5]

Yeah I have a 2K 30" sIPS screen,. I went that big at my age I can't see the text in 4K. Yeah a 32" would work. Any wider than that I would want one of those curved ones.

[kaylee]

4 hours ago, andrgl said:

(cards, subtitles, etc,) as the ranking algorithm will give you an edge

hmmm... whats this about? cards, subtitles/algorithm?

?

[Don Kotlos]

A properly sampled 2K/1080p with enough bitrate can go long way. Most often I shoot 4K to get a proper 1080p image. Thats why I wish all companies offered something similar to what GH5/s offers. An oversampled 1080p intra-frame high bitrate file. 

[TwoScoops]

8 hours ago, webrunner5 said:

Yeah I have a 2K 30" sIPS screen,. I went that big at my age I can't see the text in 4K. Yeah a 32" would work. Any wider than that I would want one of those curved ones.

I have to keep leaning forward to see my 2.7Kish screen at 37. ?

Agree with the above about youtube. Even on a HD screen if you watch the 4K youtube version it has more detail. 

[cantsin]

Cameras like the GH5s will never give you true optical 4K resolution, just a 4K pixel matrix, for the following reasons:

• With a native 4K Bayer sensor, you lose 33-40% resolution through debayering/pixel interpolation, which effectively gives you 2.5K real optical resolution. If you crop/stabilize the image, you're quickly down to 2K real optical resolution.
• 4K marks the optical resolution limit of the MFT system. Only few, very good MFT lenses resolve the necessary 12 Megapixels for true 4K resolution. (For reference, see DxOMark. Among all MFT lenses they tested, only the following resolve 12MP or more: Panasonic 20mm F1.7, Olympus 25mm F1.2 Pro, Sigma 30mm F1.4 DC DN, Panasonic Leica 25mm F1.4,  Panasonic Macro 30mm F2.8, Olympus 12mm f2.0. Most of them lose resolution on a native 4K sensor because of software geometry correction. This - strictly speaking - reduces the list to Panasonic's and Olympus 25mm lenses and Panasonic's and Sigma's 30mm lenses). This is comparable to full frame systems (with their 4x bigger sensor surface) where only very few lenses can resolve 48MP/8K.

The main advantage of shooting 4K is that it gives you true 2K - and hence contemporary digital cinema standard resolution. When you shoot FullHD/2K on a consumer/prosumer camera, you'll never get true FullHD/2K for similar reasons.

[Inazuma]

So what cameras give true 4k optical resolution?

[cantsin]

14 minutes ago, Inazuma said:

So what cameras give true 4k optical resolution?

All those on Netflix' approved camera list (since Netflix only accepts 4K for its "Netflix Originals" productions):

• ARRI Alexa LF / Alexa 65
• Blackmagic URSA Mini 4.6K / URSA Mini Pro 4.6K
• Canon C300 MK II / C500 / C700
• Panasonic VariCam 35 / LT / Pure; EVA1
• Panavision DXL 
• RED Epic / Weapon / Dragon / Monstro / Helium
• Sony Venice / F55/ F65 / FS7 / HDC-4300 / Z450

https://partnerhelp.netflixstudios.com/hc/en-us/articles/360000579527-Cameras-and-Image-Capture

[DBounce]

12 hours ago, webrunner5 said:

Are you looking at these on a TV or a 4K Monitor??

Yes 4k monitor. 

3 hours ago, cantsin said:

Cameras like the GH5s will never give you true optical 4K resolution, just a 4K pixel matrix, for the following reasons:

• With a native 4K Bayer sensor, you lose 33-40% resolution through debayering/pixel interpolation, which effectively gives you 2.5K real optical resolution. If you crop/stabilize the image, you're quickly down to 2K real optical resolution.
• 4K marks the optical resolution limit of the MFT system. Only few, very good MFT lenses resolve the necessary 12 Megapixels for true 4K resolution. (For reference, see DxOMark. Among all MFT lenses they tested, only the following resolve 12MP or more: Panasonic 20mm F1.7, Olympus 25mm F1.2 Pro, Sigma 30mm F1.4 DC DN, Panasonic Leica 25mm F1.4,  Panasonic Macro 30mm F2.8, Olympus 12mm f2.0. Most of them lose resolution on a native 4K sensor because of software geometry correction. This - strictly speaking - reduces the list to Panasonic's and Olympus 25mm lenses and Panasonic's and Sigma's 30mm lenses). This is comparable to full frame systems (with their 4x bigger sensor surface) where only very few lenses can resolve 48MP/8K.

The main advantage of shooting 4K is that it gives you true 2K - and hence contemporary digital cinema standard resolution. When you shoot FullHD/2K on a consumer/prosumer camera, you'll never get true FullHD/2K for similar reasons.

This goes a long way to explain what I saw. I honestly thought I screwed something up. Mind you, I never once before questioned the 4k image... that is, until I saw it side by side with the 2k DCI Scope image. Now I understand. 

[Damphousse]

3 hours ago, cantsin said:

All those on Netflix' approved camera list (since Netflix only accepts 4K for its "Netflix Originals" productions):

• ARRI Alexa LF / Alexa 65
• Blackmagic URSA Mini 4.6K / URSA Mini Pro 4.6K
• Canon C300 MK II / C500 / C700
• Panasonic VariCam 35 / LT / Pure; EVA1
• Panavision DXL 
• RED Epic / Weapon / Dragon / Monstro / Helium
• Sony Venice / F55/ F65 / FS7 / HDC-4300 / Z450

https://partnerhelp.netflixstudios.com/hc/en-us/articles/360000579527-Cameras-and-Image-Capture

I don't think that is correct.  All the cameras on that list that I recognize are bayer sensors.  You would have to over sample substantially to get a "true" 4k final output.

This is what netflix says...

Quote

Camera must have a true 4K sensor (equal to or greater than 3840 photosites wide).

For a bayer sensor that is techinically wrong.  I guess there is no law defining "true 4k".  But if a sensor only has 3840 photosites it can only be "true 4k" if it is a purely grayscale camera.  Otherwise all those photosites are split between red, green, and blue light capture.  You are not getting 3840 photosites capturing red light.

That is why the 1080p out of the C100 was so amazing.  It was downsampled from a 4k sensor.  The Red Monstro is a lot closer to "true 4k" than the C300 MK II.

I think Netflix just wants you to have 4k photosites.  It really is just an arbitrary line that they have drawn.  It is good for their marketing.  They aren't going to explain to the end viewer the difference between 4k photosites and "true 4k".

2 hours ago, DBounce said:

 Mind you, I never once before questioned the 4k image... that is, until I saw it side by side with the 2k DCI Scope image. Now I understand. 

Thanks for your honest and frank thread.  I haven't done a side by side experiment myself, so it is nice to see someone's genuine reaction.  All good information.

[andrgl]

17 hours ago, kaylee said:

hmmm... whats this about? cards, subtitles/algorithm?

?

TLDR; always upload in 4k if you can

 

Basically: if people aren't subscribed to you or have a direct link to your video, how is someone going to find your content?

Google needs to assign a score to YouTube content so it can determine what to show people. For your particular niche (niche being the genre of video your viewers are interested in,) you are competing with other videos.

It's safe to assume good content does well with viewers and thus your channel and videos get a better ranking.

But there's a whole chicken and egg thing at the start. If you don't have many views or videos, how do you get a better score to drive in more views? And then with very competitive niches (fashion, makeup, tech reviews, let's plays, etc) just having amazing content isn't good enough, you need an edge to beat out your competition for views.

Uploading in 4K and beyond, creating subtitles, using whatever best practices YouTube outlines for creators, all give your channel and videos just a few more points.

And that can make all the difference in getting views.

More views, more people see your work, and if you want, more ad revenue.

 

[Dave Maze]

23 hours ago, andrgl said:

If it's going to live on YouTube my belief is that you should be adapting to and using everything Google gives you (cards, subtitles, etc,) as the ranking algorithm will give you an edge. So yeah, I would upload to 4K and beyond if possible.

100%. I have had these conversations with YouTube consultants. This is true. VR, 4k at 60fps, EVERYTHING YouTube is offering you need to prioritize that for analytics advantage.