Is anyone else interested in a screen-less (small) external recorder?

[kye]

On 9/12/2021 at 11:55 PM, Video Hummus said:

Never did I say you needed to buy an R5 or that even R5 was superior in anyway. Not sure what you are reading.

Again, no I did not. I suggested today’s cameras and future ones will make external recorders less useful in the IQ department which is good because I don’t like the concept of external recorders at all and find latest internal codecs get me 95% or more there. I would welcome any smallish mirrorless camera with internal ProRes. 

Come on, like a codec is the only measure of the “organic look”. I think we both know it’s far more complicated (and subjective!) than that. 

Actually, I'm a few steps ahead.

By saying "I'll do a blind test of codecs with the R5" what you're really saying is that you don't think that there's much difference between the codecs (no-one is suggesting blind tests between GoPros and Alexas).  The followup implication is that if there is little difference between them on your R5 then either:

1. there is little difference between codecs and my perception is flawed somehow
2. there is a difference between them on cheaper cameras but not on expensive cameras

Either way, both are an argument against the idea of adding an external recorder to the existing equipment I have.

In terms of the codec, you're right that it's one element of the 'look' but what you might not know is that I've been doing testing offline over the last two years and have exampled many variables.  I simply stopped posting them here because of the increasingly ridiculous and trolling responses from various users.

I own several cameras of note:

• Canon 700D with ML installed (which can do 14-bit RAW up to about 1700px wide and has Canon colour science)
• BMPCC and BMMCC (which can do 1080p RAW and Prores in various flavours and has excellent colour science)
• GH5 (which downsamples from ~5.2K, does very high-quality h264, and has lacklustre colour science)
• Various MFT, pocket, and action cameras (which shoot low-quality h264)

I've extrapolated quite a lot from comparing and contrasting various things from these cameras.  Over literally dozens of tests conducted in meticulous controlled conditions, often using identical lenses and other equivalencies, I've started noticing patterns.

I also watch a lot of streaming content, which is often shot on high-end cameras (RED, ARRI, Sony Cine cameras), but is heavily compressed / sharpened as part of distribution.

A few things that have gradually emerged from all this is:

• DR
  DR matters as not only do things clipping look awful, but DR is literally measured by setting a threshold for the noise floor - so more DR = lower noise floor, and that means cleaner shadows
• Colour Science
  RED and ARRI footage is immediately noticeable for looking delightful even at 480p setting on YouTube, which has poor resolution and poor bitrate, I suspect colour science (and DR) are what shines through in this situation, where resolution, sharpness, etc will be obscured
• Codec and processing at capture
  If I review a bunch of footage captured by me at 200Mbps h264 and compare it to footage shot at 1000Mbps+ but streamed at 17Mbps (YT4K) the YT looks nicer.  If I compare footage I shot on a cheaper camera (for example my Panasonic GF3) at 17Mbps 1080p and compare it to ARRI footage streamed at 2.5Mbps (YT1080p) then the YT looks better and less artefacts

I've spent literally hundreds of hours on this stuff, and the more I look the more that I narrow things down and the codec is a big one.  You're right that h264 probably isn't much of a problem, but when it comes with very low bitrates and lots of sharpening and NR (which compound the artefacts of the h264 compression algorithm) that's where the issues start, because it's a package deal.

[kye]

As an addendum to the above, this article has a number of very interesting points:

https://ymcinema.com/2020/06/08/reds-baby-dragon-komodo-an-action-cinema-camera/

Of particular note is the fact that GoPro footage is noticed by audiences after only 20 frames.  Combine this with the fact that GoPros shoot at 4K at up to 100Mbps, but cinemas are often still in 2K.

The article talks about rolling shutter and DR of the GoPro being limiting factors, but when I look at GoPro footage (even the latest v10) it looks awful even in low-DR situations without much movement.  Of course, noticing the difference between a GoPro and cinema camera in low-DR low-movement situations will take more than 20 frames, if we're talking about it as a main camera, the differences are obvious.

But let's think about this further.

A GoPro will still be visibly inferior under the following situations:

• Low-DR situations
  If the GoPro is still inferior in low-DR situations then it's not the DR driving the difference
• High-motion
  Once again, GoPro still has it's poor look in low-motion scenes 
• Deep DOF situations
  Most action scenes in movies will have deep DOF used by the cinema cameras to show the relationships of all the moving objects, so this isn't a differentiating factor
• FOV
  This is potentially a factor, with the GoPro look being very wide-angle, however I have modified an action camera (not a GoPro but the similarities are there) with longer focal lenses (multiple) and they do stop looking like action cameras but don't stop looking like cheap/nasty cameras.  A pocket camera with long zoom also doesn't get 'nicer' by zooming.  Also, I have attached wide lenses (GoPro wide equivalent) to my GH5 and it doesn't suddenly look like GoPro footage.
• Resolution
  In theory, downscaling an image increases its quality, so having a 4K image put onto a 2K timeline should give it the advantage over a 2K cinema camera, but no, no it doesn't.

So what is left?

Colour science remains, image processing remains, codec remains, bitrate remains, bit-depth remains.

Colour science is potentially a thing, but that "GoPros are noticed after 20 frames" was in the final movie that has been graded by a world-class colourist who will be pulling out every trick they know to hide this particular shot.  Colourists can't work miracles but they sure can work magic, so I'm going to suggest colour science isn't a main thing.

Now bitrate, bit depth, and codec.  The A7S2 (with its 4K 8-bit 100Mbps h264) is a parallel here, with wedding film-makers generating higher quality footage than GoPro can, but I suspect that an A7S2 doesn't cut well with cinema cameras, and the A7S3 made large improvements in this department, for precisely this reason.

That leaves image processing.  I suspect any professional camera is likely to process the image with much higher quality than a GoPro, and with this I include the quality of the compression itself, which we know is a highly variable thing.

It is probably a mixture of all these things, but I'd suggest it's likely image processing first, quality of compression second, bitrate third, codec fourth, colour science fifth, bit-depth sixth.  However, anyone offering Prores will automatically deliver much improved processing, bitrate, and bit-depth, and I suspect that encoding Prores is also much less computationally demanding so can also be done at a higher level of quality too.