Youtube 4K quality is so poor you might as well shoot 1080p

[kye]

Yes, this is a big claim.

Allow me to explain how I came to this conclusion, and prove it to you.

First off, the evidence.  This is a video containing 5 compositions that were each shot with either 2K, 2K (and processed in post), or 4K.

It should be easy to tell the difference - so see if you can tell!

Not so obvious is it....

In the video I encourage you to download the video and pixel peep, please do.  

I have pixel peeped the shots directly on the timeline in Resolve, in the output file I uploaded and the YouTube file I downloaded.  It took me zooming in to 200% and finding the place in the video with the finest details, to be able to see the differences, and I know which is which!

Now, the details, to show that the test is valid.

I shot each scene with the GH5 and 42.5mm Voigtlander (stopped down several stops) in either:

• 4K 422 10-bit ALL-I 400Mbps mode
• 2K 422 10-bit ALL-I 200Mbps mode (with Resolves Super Scale)
• 2K 422 10-bit ALL-I 200Mbps mode (with no processing)

I exported the timeline to an RGB Uncompressed 10-bit file (65GB - 6,500Mbps!!) and compressed that file using ffmpeg to a 4K 10-bit 422 IPB 225Mbps h265 file, which was then uploaded.  I tried uploading an 10-bit 422 ALL-I file but YT only interpreted that as 1080p.

YouTube then compressed that 225Mbps file to the pathetic 10.42Mbps file you see if you watch the above YT video in 4K.

As far as I can tell, there's nothing I could have done differently to get a higher quality result out of YouTube.

But what about other platforms or delivery methods?

This is just YT.  If you pay for Vimeo, or deliver via any other mechanism that has a higher bitrate than YT this thread should make you feel better about that!

But 4K is useful for things like cropping in post!

Yes, and this test doesn't apply to doing that.  I've tested how much you can upscale an image without it being visible in another thread / test (spoiler, it's something like 150%) but let's leave that aside for now.

But YouTube supports 6K and 8K and ......

Sure, and when people start watching YouTube with their 6K and 8K TVs then they'll start to benefit from that.  Until then, they're getting the 10Mbps file above.

Are you saying that a 4K camera is not required?

There are advantages to having a >1080p sensor.  Downscaled video is much nicer, and there are times when shooting in 4K or higher can have advantages.....

So, when IS it useful to shoot in >1080p?

Lots of times:

• Cropping (significantly) in post
• Overcapture for things like stabilisation
• If your cameras 1080p isn't that great (which is most cameras TBH)
• etc.

I have nothing against 4K or higher resolutions except that I think most people are making their life harder than it should be thinking that they're somehow getting better results when for many people it simply doesn't matter if they shoot in 4K or not.  This was me.  I fell for the hype, and have gradually been doing tests like this to actually see for myself what is true and what matters - rather than just believing the marketing hype from the camera industry.

Ultimately, the lesson here is that what matters is that you publish in 4K, not that you shoot in it.

Let the questions and comments (and flame wars from resolution fanboys and fangirls) begin...

[hyalinejim]

Love me some 1080p!

A fortnight ago, I shot 5 hrs of interviews for a 5 minute corporate video. 1080p 100mbps all the way. I do lots of jobs like this and like to keep everything I shoot in case a client wants to use it again. I hate when hard drives fill up and I have to deal with offloading projects to external backup. Give me clean 1080p every day of the week!

I shudder when I see specs like "8k, 960mbps". Where is all this footage gonna go?

[SteveV4D]

3 hours ago, kye said:

 

I have nothing against 4K or higher resolutions except that I think most people are making their life harder than it should be thinking that they're somehow getting better results when for many people it simply doesn't matter if they shoot in 4K or not.  This was me.  I fell for the hype, and have gradually been doing tests like this to actually see for myself what is true and what matters - rather than just believing the marketing hype from the camera industry.

I don't think its that big a claim to day that Youtube is rubbish and that quality isn't all that great uploading 4K over 1080p.  Aside from small videos, most of what I deliver is 1080p.

But to say I'm making my life hard shooting 4K - nah.  I'm loving it.  I shoot nearly always 4K BRAW and the pleasure editing these files compared to H264 and even worse, those AVCHD files that for some reason struggled to playback in Premiere even though 4K H264 were okay.  

It use to be that after I graded the footage, I had to render to see any benefits and that was just a simple grade.  Now I perform complex grades in Resolve that playback fine on BRAW, whilst stuttering with H264.  

Probably about 95% of my work I am doing things where 4K is an advantage.  And should the Industry move closer to 8K, 4K will upscale better than 1080p.  

I doubt I will ever go back to shooting 1080p for any project.  Most modern cameras now don't give much thought to 1080p anyway.  I wonder how long it'll be before it is missing altogether from new cameras.

[kye]

17 minutes ago, SteveV4D said:

And should the Industry move closer to 8K, 4K will upscale better than 1080p.  

Seeing the results of my tests, I the logic will be transitive.  ie, when people start watching in 8K, then it might be time to start shooting in 4K.  

Of course, people watching in 8K probably won't be watching on screens twice the diagonal size, so it will actually matter much less. 

19 minutes ago, SteveV4D said:

I doubt I will ever go back to shooting 1080p for any project.  Most modern cameras now don't give much thought to 1080p anyway.  I wonder how long it'll be before it is missing altogether from new cameras.

Agreed.  One of the things that is interesting to most (but not you as you've moved to shooting RAW) is that to have a 4K pipeline your computer has to decode and process 4 times the pixels.  By putting 4K files on a 1080p timeline the computer would still have to decode 4K but then would only have to process 1080p.  Anyone who is wanting really nice intermediaries can shoot 4K compressed, and then render 1080p proxies in an ALL-I format, then they only need to decode 1080p and process 1080p.

I used to think that somehow the 4K quality was visible and that if I had some sort of 1080p processing along the way then it would be like I rendered to 240p and then was exporting at 4K again but the weak link would screw everything up.

Turns out that the weak link is actually after your entire workflow, so no point keeping all the quality (and paying for the processing of it) in order to just have it crunched after you have delivered it.

I remember when 4K was a huge deal and people were beside themselves about how much storage and processing power it would take for 4K.   Hell, I remember buying the first 1080p panel that cost less than a reliable second-hand car, and it was great, but it would be two computer upgrade cycles before my computer was powerful enough to play any 1080p videos!

Now we've gotten computers powerful enough to play 4K we naturally assume that it's easy and therefore not a requirement.  Then we lump on all these effects and processing in post and spend thousands of dollars on something that we "don't give much thought to".

Turns out it all doesn't matter, for YT that is.

[Andrew Reid]

I can tell the difference between 4K YouTube and 1080p.

Just not in Safari!

[andrgl]

@kye Sitting infront of a 4K 43" monitor, the difference in resolution between clips is easy to spot.🤷‍♂️

I could also tell when clips were sharpened because of the microcontrast haloing.

 

The order of clips was annoying, had to rearrange so they were back to back, not sequential.

[kye]

58 minutes ago, andrgl said:

@kye Sitting infront of a 4K 43" monitor, the difference in resolution between clips is easy to spot.🤷‍♂️

I could also tell when clips were sharpened because of the microcontrast haloing.

And the answer is....?

58 minutes ago, andrgl said:

The order of clips was annoying, had to rearrange so they were back to back, not sequential.

If you can't tell without a back to back test, then you won't be able to tell when someone shoots something in one resolution and you never get to see it in another.

[kye]

2 hours ago, Andrew Reid said:

I can tell the difference between 4K YouTube and 1080p.

Just not in Safari!

Between 4K YouTube and 1080p YouTube?  Hell yeah - 4K is 10Mbps and 1080p is what...  0.1? 😂😂😂

lol about Safari.  Maybe I should do a blind test for safari viewers between 360p and 1080p!

I literally use safari for all browsing except media consumption 🙂 

[kye]

1 hour ago, andrgl said:

The order of clips was annoying, had to rearrange so they were back to back, not sequential.

Actually, for it to be a real test, I should have shot different scenes altogether.

[Ken Ross]

3 hours ago, Andrew Reid said:

I can tell the difference between 4K YouTube and 1080p.

Just not in Safari!

As can I...and I'd bet many arrive at the 'no difference' conclusion between 4K & HD while watching on Safari. Many are still unaware that Safari won't do 4K on YouTube.

[monk191]

Apparently, Safari will finally support 4K YouTube in the next version of OS X releasing soon. (And Apple TV too)

[markr041]

I can easily tell the difference between 4K and 1080 on YouTube. I can even tell the difference between 2880P (5K) and 4K on YouTube, even on 1080 screens. I don't use Safari.

[majoraxis]

I'm on a 5k iMac viewing in Safari and I left the playback resolution in auto which I believe defaulted to 1080p.  I thought setting B on the video looked best. I noted a difference around the edges of the image elements being that they seemed to have more resolution and less offensive artifacts.

What was setting B?

If setting B was 2k then that will same me upload time and bandwidth.

Thanks!

[scotchtape]

I think C is 4K, the rest, I don't know, nor would I expect to see a big difference between the two modes.

4K works for me! 

[seanzzxx]

People who confidently claim to immediately spot the difference but don't provide examples, please feel free to prove your case.

[KnightsFan]

For me A > B > C. C looks the worst to me but mainly because that shot at 1:08 looks digitally sharpened which was visible in a casual viewing. I wouldn't have noticed any other differences without watching closely. I'm not sure how matched your focus was between shots but in the first composition A seems a little clearer than B which is a little clearer than C, could easily be focus being off by just a hair though.

[Andrew Reid]

It's also shot type dependant.

Close-ups of plants, oranges, etc. don't demand as much resolution as a wide angle shot of a landscape at infinity focus.

[kye]

1 hour ago, KnightsFan said:

For me A > B > C. C looks the worst to me but mainly because that shot at 1:08 looks digitally sharpened which was visible in a casual viewing. I wouldn't have noticed any other differences without watching closely. I'm not sure how matched your focus was between shots but in the first composition A seems a little clearer than B which is a little clearer than C, could easily be focus being off by just a hair though.

I focused by focusing at f0.95, then stopping down several stops.  I think I shot at either f2.8 or f4 (stopped down 3-4 stops).  I then shot in whatever mode the camera was in 4K / 2K, if the camera was in 2K then I would wait a few seconds then shoot again, then change to the other mode and shoot.  

I shot handheld standing up, so slight focus changes are possible.

In post I used a random number generator to decide which of the 2K shots got Super Scaled and which was kept unprocessed, and also on the sequence of how the 2K SS / 2K / 4K appeared in the video.

I've done tests comparing how much lenses sharpen up when stopping down (closing down by 2-3 stops is normally close to their sharpest) and I compared the Voiglander 42.5mm to the Helios 58mm + 0.7x SB (41mm) and the Konica Hexanon 40mm and all were similar sharpness when stopped down, on MFT at least.

Once again, if it's only visible by direct comparison of identical shots then it won't matter because I don't get the option of watching a YouTube video shot in 4K and then watching the same video shot in 2K and then up-scaled.

[kye]

1 hour ago, Andrew Reid said:

It's also shot type dependant.

Close-ups of plants, oranges, etc. don't demand as much resolution as a wide angle shot of a landscape at infinity focus.

Agreed.  I should do a follow-up with different scenes, but including some landscapes at infinity focus.

Having a composition where lots of stuff is out of focus really helps any compression algorithm along, so having lots of detail is much harder.

Would you suggest a shot with lots of stuff moving?  That's not that easy to find, although I guess the ocean would be pretty good for lots of random motion at a far focus distance.

[KnightsFan]

With the handheld footage and moving subjects focus is lost quite a bit. The shot of the orange for example I see no difference between all 3 since the orange is moving in and out of focus. But on that first shot of the plant, which is relatively still, A is the clear winner.

B and C are pretty close, but I do think that one shot has C losing ground.

32 minutes ago, kye said:

Would you suggest a shot with lots of stuff moving?  That's not that easy to find, although I guess the ocean would be pretty good for lots of random motion at a far focus distance.

I'd suggest shots that are similar to what you actually use since those will be most useful. I'm on a tripod most of the time, which is where you'll see the most difference between 4K and HD especially with IPB compression.