KnightsFan

The a7 is old. This is a 2019 camera we are talking about. I assume it will have 4k and high frame rate of some sort, etc, much different from a years-old camera having dropped in price to $800

I wonder what the Z5 will cheap out on. $1200 for full frame is nuts.

Do you mean the 20xx line or the 10 series? transcoding 4k h265 on my 1080 using hardware acceleration is lightning fast, though editing isnt as fluid.

Similar price, both will be used by vloggers, hobbyists, students, and professionals for a variety of purposes, and both will end up in those "$1000 camera vs Arri Alexa" videos. Both have some features i like, and lack some that i find essential. Considering the price, i find a comparison between this and the p4k more apt than say a comparison between this and an a7s3, or between the p4k and an ursa mini. In both of those cases, one option is simply out of my budget. If you want to tell people which cameras can be compared, youve got a lot of internet to police.

This seems like a decent budget camera. Nothing groundbreaking, true, but it has oversampled 4k, and HD 120 all for under $1k brand new. I would like to see HEVC, but Sony has never been class leading in the codec department so no surprises there. I would compare it to a P4K with half the video goodies removed, and some decent photo and AF features added.

It might be a year or two, but in the meantime we can use proxies, the way we did when 4k just started. I dont think computing power will be a big deal in the long term. It seems to me that 8k is happening much smoother and cheaper than 4k did. Thats not to say that 8k is super important, but i think the tech is shaping up nicely.

It would not work. You can only adapt lenses to mounts with a shorter flange distance. Canon FD < Canon EF < Nikon F. That's one reason FD lenses tend to be so cheap on the used market: they can't be used on modern Canon cameras. Well technically there are some adapters that use corrective optics to allow FD lenses on EF cameras, but they are generally expensive and reduce the quality considerably.

At the moment just a 50mm f1.8 SC that I found in a local thrift store a few years ago for $10. Sometimes I think I should try to sell it, but it's my sharpest 50 and so I use it for VFX-related photography where I really need detail. It's such a beautiful lens, but it's just too much hassle to use for standard video work.

I stack adapters. All of my Nikon lenses have a "permanent" Nikon F -> Canon EF adapter attached. Stacking cheap adapters can result in more play overall. However, most F/EF adapters can actually be tightened with a flathead screwdriver so there is no added play. (They aren't permanent really, but I haven't taken the adapters off in years) Then, I use a Canon EF to Samsung NX adapter that lives on my NX1. I basically just treat my NX1 as an EF camera now, as @kye suggested. If you regularly switch cameras and lenses, I highly recommend settling on one standard mount and converting everything to that. I've used my same lens set on 5D3, NX1, XT3, Ursa mini, and others. And sometimes on shoots I've used Canon, Rokinon EF or Zeiss ZF lenses alongside mine. So settling on EF as a standard has really made my life easier. Unfortunately, that means that my beautiful Konica and FD mount lenses haven't seen use in a long time.

Exactly true, and the reason why I bought a used NX1! However, this topic is comparing the P4K and Z6, so I brought up a point that hadn't been discussed, yet would factor into my decision between these two specific cameras. The point is not that you should get a Z6 for sightseeing. My point is that if you get a camera, the Z6 has some uses that the P4K falls short in. Similarly, if your work requires 4k60 then the P4K is clearly the better choice. There are any number of arguments and factors, I brought up one that is important to me. Right, and that's perfectly reasonable. However, as shown in this example, you can use H.264 and save 90% of the space, while still retaining more information. Not compared to H.264! In this example, ProRes LT comes out to 57 MB, with a PSNR of 33.56. H.264 has a PSNR of 33.35 at just 3 MB, roughly 5% of the size. I mention this because PSNR has been brought up already, despite not being a great metric for visual quality. So subjectively, I see that even LT has mildly less visual detail at a 1:1 crop. ProRes is great for being widely standardized and accepted, and it plays back smoothly, but space efficiency really isn't its forte.

Just downloaded and tried PotPlayer, still no green tint on my end. I have no idea what could be causing it for you, but you may want to try changing your renderer in PotPlayer, or at least verify whether it happens with every H.264 file or just these. Of course you can use it any way you want, I just think there are better options available. And, to get back on topic, the whole reason I brought it up was that I would rather have a Z6 on a casual sightseeing trip than a Pocket 4K. Both for photos and videos. I use my camera for paid gigs, but also for non-pro stuff. So that's a genuine consideration I need to make before buying. I have yet to see a place where ProRes proxy is better than H264 with the same bitrate. As I said, YouTube comparisons between completely different cameras don't really show anything about the acquisition codec. But yeah, if it works for you then use it. That's fair. "Terrible" in an absolute sense might be harsh, but in my experience, it's not as good as H.264. Same! I guess I've derailed the topic enough so I'll leave the files as they stand and let people draw their own conclusions.

It seems that somehow you have added a green tint that I cannot reproduce. That's why I asked which software you used, so I could check it on my end. I've tried my clips in VLC, MPC-HC, and DaVinci Resolve and I can't get the green tint. What do you mean? The image I posted is a 1:1 crop, it was not blown up. You can see the exact same artifacts in the image you posted. I posted four total samples, one ProRes Proxy and three variants of H.264. What further H.264 comparison do you want? Feel free to send me clips to compare. As I showed, even the H.264 All-Intra variant is visually and mathematically better, so motion shouldn't make any difference. This is absolutely true. However, all of the samples I posted were directly transcoded from the original 10 bit 4k file with no edits or effects. I didn't push either one. Again, feel free to send Raw or Uncompressed source files in the scenarios of your choosing and I'm happy to run the test again.

Ran PSNR analysis against the original file. ProRes Proxy avg: 31.52 H264 (same bitrate) avg: 33.79 H264 (<1/10 bitrate) avg: 33.35 H264 (All Intra) avg: 31.73 The H.264 files are all mathematically closer to the original file, in addition to being visually closer. Even the All Intra version!

@Emanuel There is no greenish tint in Resolve, though I see it in your screenshots. It must be an artifact from your player. Are you using VLC? EDIT: Yeah, definitely an artifact from your player/screen capture software. Your screenshot is significantly greener than the original file when shown side by side in Resolve. Camera in a tripod, I think. It's an old test clip I found in my PC's recycling bin. I was testing settings on my friend's XT3. Even the All I version of H264 retained more detail and didn't have those weird blurry bits moving about. And for casual video, acquisition and delivery happen on the same clip. I'm not going to transcode just to upload to Facebook. And ProRes Proxy wasn't designed for either, it's designed as proxy. Can't tell if you're joking about ProRes being even remotely close? I mean I guess if the artifacts below are acceptable to you then so be it. I mean just look at the jagged edges on the roof and the bands of blurriness, which move up and down as you play the video. H.264 looks clean and consistent. This is a 100% crop from ProRes Proxy. I'm not even zooming in at all. Can't see how this is at all acceptable for pro use except as proxies.

AND... just for fun, I encoded using H264 All-Intra. It was an 8 second encode (fastest of all the tests, actually--maybe I should drop encoding speed to 20 seconds and see if it improves). I think this is a subjective one. This version clearly has a noise overlaid, but resolves more detail than ProRes Proxy by a good margin. Noise vs reduced resolution? Very tough choice. Both are pretty bad. Maybe for our scenario--family videos--ProRes is better since people might find the flicker distracting? This is a worst case for H.264. 99% of scenes will benefit significantly from P and B frames. https://drive.google.com/open?id=1lie32Rfmezd0eLH_IrQ5TQbUA9MsPJ85

@Emanuel I might do some more scientific tests later, but here's an example. The source was a 10 bit 4K XT3 clip shot at 200 Mbps. I transcoded using ffmpeg. To make it more even, I lowered the CRF on the H.264 until it encoded at over twice the speed of the ProRes version. Hardware encoding was off. This means that the H.264 actually took less processing power to encode. We could actually increase the quality of the H.264 clips without adding any extra data simply by using a slower setting. ProRes Proxy: https://drive.google.com/open?id=1s8jpbgEvBQdduMul90kdDovQEU-xeJ0e H.264 (same bitrate): https://drive.google.com/open?id=1G8caHjIW--qLSmNNoLSkaBl2g2sPDU0j H.264 (<1/10 bitrate): https://drive.google.com/open?id=1fIg87fLa0V0KvCnGksRoHEGEWgROadKz Download and watch them at 100 zoom. I think it's clear that both H.264 files look better than ProRes Proxy in this scenario. I also notice that when watching in the Google Drive viewer, I can't see much difference (thanks, YouTube compression!), although it is night and day when downloaded and played in the app of your choice. Do you disagree? If you think the difference is an in-camera encoder being worse, just imagine if a C100 looked as bad as that ProRes Proxy shot. And remember that in this case H.264 encoded in under half the time. Naturally there are MANY problems with this test, but I didn't have much time today. Obviously H.264 benefits from it being a mostly static shot. (though to be fair, the ProRes won't get any BETTER with more motion, it's just that H.264 will get worse). It would be interesting to also test some scenes with a LOT of motion. A handheld walk through the woods perhaps, or even a video that cuts every single frame to a different scene. If you can source some 4k RAW clips or uncompressed RGB 444 HD clips, I'm happy to do more tests in conditions of your choosing.

@Emanuel ideally we could find someone to record on both an atomos and internally on really any camera. In 1080p, prores proxy is about 35 Mbps so a f3 would be great. In 4k we could compare a gh5s in 150 Mbps mode with prores proxy. I dont have an external recorder, so on my own i am limited to finding some raw or uncompressed footage and using ffmpeg. Obviously thats using an overpowered encoder, unfortunately, though i can lower the encoding speed. The benfit of doing it with ffmpeg, though, is that we can run psnr tests if you are interested, though psnr is a really bad metric for video quality. From wikipedia: "Although a higher PSNR generally indicates that the reconstruction is of higher quality, in some cases it may not. One has to be extremely careful with the range of validity of this metric; it is only conclusively valid when it is used to compare results from the same codec (or codec type) and same content.[8][9] Generally, PSNR has been shown to perform poorly compared to other quality metrics when it comes to estimating the quality of images and particularly videos as perceived by humans." And you are right that we are mainly interested in a specific context. My original context was casual use: snapping a family video, or uploading to facebook with little to no editing. Ive actually seen both those videos before. The first test is flawed because the canon is a completely different processing pipeline. The codec is one of many variables. Both tests are flawed as they were uploadrd to you tube, which is a significantly lower bitrate than either of the two original clips. So unless you are specifically trying to decide which looks better on youtube, it is pretty useless to compare codecs in a youtube video.

@Emanuel maybe i should do another blind codec test for everyone comparing h264 with prores proxy?

@androidlad jeremy says in the video that nikon hasnt decided whether to use 10 or 12 bit, or whether to use linear or log yet. According to thr prores raw pdf, it "decodes to linear" which means that it could be stored in log. In any case, the format doesnt know which the bits mean. Saying prores raw is linear is like saying hdmi carries a yuv signal... atomos and nikon can put whatever they want into that signal. @Eric Calabros seems right. At the moment they are recording 12 bit linear, but the specs of what they will release to the public are not finalized. Not that i take the 10 bit as a fact either, its still up in the air.

"Z CAM E2 is formally ProRes licensed!" -Kinson Next firmware update will unlock it automatically; in the meantime they posted a method to unlock ProRes manually. Unlike the internal H.265, which is 10 bit 4:2:0, internal ProRes on the E2 has 4:2:2 subsampling.

Yup. You'd have to use a very poor H.264 encoder if it looks as bad as ProRes Proxy at the same bitrate. Again, not talking about serious production here. I'm saying for people who want to use their camera for casual purposes, the presence of lower bitrate video is a benefit, while obviously the Z6 with an Atomos has ProRes as well.

ProRes Proxy looks pretty terrible compared to H.264 at a similar bitrate. LT is 3x the size. ProRes really just isn't a great family of codecs for casual videos.

Let's not forget that the Z6 can shoot in standard H.264 as well! That's a huge benefit for hobbyists or enthusiasts who want high fidelity for projects, but also use their camera for casual purposes like family vacations.

I never said uncompressed, but Raw should be mathematically lossless to be a legitimate Raw format. It should also have a linear gamma, little to no digital processing, and no debayering if originating from a bayer sensor. Any video format can be treated as a "digital negative" and store information that can be recovered later. Most log curves are designed to do just that. If I make a video format that simply takes the bayer data from a sensor, pretend that every group of 3 pixels is a single RGB pixel, then apply ordinary ProRes compression to the entire thing, that's certainly not RAW at all. I suspect that is very nearly what ProRes RAW does. You can essentially change the white balance on any video if you transform it into linear gamma before doing so. The only hindrance is that most cameras don't store the original white balance metadata in non-raw formats. As long as ProRes RAW retains all the information from pixels sufficiently, it should be just as flexible as any of the other Raw-lite formats for white balance adjustments. It does seem silly that it is not builtin, though.

I agree with @GiM_6x. We should put some effort into using words correctly. Its the reason netflix wont allow upscaled alexa footage to be considered 4k. Prores raw, redcode raw, bmraw, are all fine formats. I dont have an issue with them as formats, but it is marketing bs to call them "RAW." Its not a matter of whats "needed" for a given production, its quite simply a matter of using correct terminology. Its like putting an 8 bit stream into a 10 bit container and calling it 10 bit.