The Pocket 4K footage we've all been waiting for

Lol! No, of course not, it's for effect. That last one looks rather nice (is that restored?). The tool is to give an image to shape. So, my point is 100% valid, the better the camera can capture professionally, the more options you have to do things. But let's hook over to the title footage of the TV series, Becker. It's brilliant, I like it very much. Shot on a old camera with not much lowlight performance, resolution etc etc etc.
But from what people hete are saying, It's almost like saying we should be happy if people shot everything like that. Please, no, save our eyeballs. A pocket 4k that shot only like that, maybe a very poor seller, except maybe in the 18th century before or after the impressionists. My favourite oreintalist's paintings were rather good. He was largely after David, and opposed the impressionists fur many years, eventually becoming a sculptor. There were some romantic fantasy artist after this that refined the finess of the technique further before the hyper/super realists etc. Those paintings were exquisite.

But people get Art confused with that which you inflict on others compared to what others will appreciate. It's like brain washing. Frankly, the packing room prize is likely to go to paintings with better lighting.

If you look at Australian art before the arrival of European artists from the wars last century, it was rather strikingly beautiful, idealistic, imperial in an.australian ideal, but then demented neurotic after the arrivals. Of course, these things would have been in the underground before hand, but we're not promoted. But as the famous Australian art tv series title said "The shock of the new". People are swept away by interest in the new and shocking, and unfortunately, when artsists are in charge of that, things go awry. They don't see like normal people see, and try to convince the wider public of what is art and loose them (after whatever ride in the meantime). They see past what other's may see as art, as something that pleases themselves. Just unfortunate. But explains so many film grades I have seen since the Red revolution.

I feel like I have to explain often to people that don't understand shooting video: "natural" lighting usually looks terrible, and planned lighting that looks off in real life looks great on camera.

There's a reason there is a lighting style used often in film that is literally called "Rembrandt lighting".

Maybe I do understand, and it's not just one school (/FAD), particularly overused on sensors that make it look flawed. The problem is you not only get harsh clipping, but the colorisation and tones of the face Jump unnaturally. A better sensor, similar to those photographs quoted, can reduce that without makeup, allowing you to light details and bumps of the skin, without looking like somebody that hangs around chimney sweeps.

Anyway, nearly 4am here guys, I was ready to go to bed around 1am before the artinquisition. Anyway, if you light properly you should get 15 stop+ like results. For one person bands it should be good enough without that if they are careful on post and watch what they aim it at.

I should point out here, the jumps on skin tone I mentioned are minor, that's why I think it can be dealt with, compared to older sensors less than 11-12 stop.

But let's hook over to the title footage of the TV series, Becker. It's brilliant, I like it very much. Shot on a old camera with not much lowlight performance, resolution etc etc etc.

The Becker...title sequence. You mean this?

Yeah, I think I'm seeing the basis for our disagreement, if your vision of a good image has a heavy posterize effect on...all of it.

Becker was a sitcom. You do know that sitcoms light everything, right? Specifically, multicam sitcoms use extremely flat lighting - like Becker, Cheers, Big Bang Theory, Roseanne, etc. If that's your favorite look, more power to you.

I prefer cinematic lighting over flat lighting, but that has nothing to do with your camera. And it doesn't matter if you light to 15+ stops if your camera can only record 8-9 stops of exposure (like old DSLRs, 5dmkii, for example).

Yet another attempt, less saturated bokeh this time.

Davinci_grade_15_02.23.19.jpg (103.75 KiB) Viewed 3689 times

Fascinating thread. What a lot to think about!


Sent from my iPhone using Tapatalk

But let's hook over to the title footage of the TV series, Becker. It's brilliant, I like it very much. Shot on a old camera with not much lowlight performance, resolution etc etc etc.

The Becker...title sequence. You mean this?

Yeah, I think I'm seeing the basis for our disagreement, if your vision of a good image has a heavy posterize effect on...all of it.

Becker was a sitcom. You do know that sitcoms light everything, right? Specifically, multicam sitcoms use extremely flat lighting - like Becker, Cheers, Big Bang Theory, Roseanne, etc. If that's your favorite look, more power to you.

I prefer cinematic lighting over flat lighting, but that has nothing to do with your camera. And it doesn't matter if you light to 15+ stops if your camera can only record 8-9 stops of exposure (like old DSLRs, 5dmkii, for example).

Sorry, my mistake Michael. The end credits, the night time shots, not the actual show itself. As you can see on the other thread, I'm a fan of many looks, flat or cinematic. Just that the extra stops make the cinematic ones better, and you are going to have to control lighting more to make it look good on a camera with 8-9 stops you mentioned This means controlling fill so as not to overdo it, and to even out colorisation issues. You see how in those photos posted, while they had the look you were looking for, the skin between the shadows is that bed smoothly and beautiful, it the rise and fall of darkness which is being plaid with, not making the person look like a zombie or severely sick, which I not happy with. Indeed, the nose of one ofmthe paintings shoes a lack of detail, but even if the shot/file of it is not over exposed, the finer detail and colorisation in that highlight area could have faded and even dropped off over the centuries. I'm sorry it was a bit amusing to me guys, bit it was bit like being a knife to a duel. It just had issues. The photographs though, were good/fresh examples.

Have a look at the 22nd second of that clip.

...and a grade of RAW footage. Added some motion blur as it is way to sharp to my taste.

Davinci_grade_15_ 11.18.47.jpg (190.68 KiB) Viewed 3643 times

Most of the time, I think DNG's speed is gated by disk performance rather than CPU or GPU.

Nope- nothing to do with disk or debayering. I'm talking pure CPU needs.
For example: it just manages 24fps on my 4 cores Macbook Pro, where same size Cineforma RAW (which goes through the same debayering process) can go way higher. You can even lower debayering to 1/4 resolution and it doesn't change anything (GPU is only at 30% for full res debayering, so it explains it). Problem is CPU which is close to 100%. DNG is for whatever reason so slow. Take same clip as final video- ProResHQ etc. and even this plays way easier. This is not normal as DNG should be way easier to decode than any intermediate codec (RAW is just a monochrome image with 1 channel). It's like 3 or more times slower than it should be.

Why not? BMD and Sony have both been doing that sort of thing for quite a while, I'm fairly certain that they've ironed out their glitches by now. Plus, it's a back-side illuminated sensor, so it's inherently more sensitive than its predecessors just because of having no circuitry in front of the photosites.

I'm talking about real usable data. I don't think that you will get much advantage from 12bit RAW over 10bit ProResHQ when it comes to headroom for adjustments. Point is that 10bit ProResHQ isn't really a problem as a recording format. It doesn't really limit what camera can actually record (maybe tiny bit).

I'm talking about real usable data. I don't think that you will get much advantage from 12bit RAW over 10bit ProResHQ when it comes to headroom for adjustments. Point is that 10bit ProResHQ isn't really a problem as a recording format. It doesn't really limit what camera can actually record (maybe tiny bit).

Sadly there is a visual difference side by side between RAW and ProresHQ.. even so. 10bit ProresHQ is good enough for many professional productions and you do get a lot of headroom in post.

The biggest selling point, IMHO, is that it’s a huge difference from a 8bit h264/h265 compare to a proresHQ10bit footage.. And BMPCC4k doesn’t seem to have the baked in NR/sharpening you get in a 10bit GH5/GH5S footage...

Not talking about visual difference, but amount of data which I can "play with". If there would be very clear 12bit worth of data in RAW then 10bit in ProRes would "reduce" this substantially.
Main "visual difference" is most likely due to better debayering in Resolve. In camera processing is always a compromise and RAW is definitely best format to start with.
My question is: does 10bit recoding format really limits camera possibilities (forget about fact ProRes is lightly compressed and 4:2:2 etc)?

My question is: does 10bit recoding format really limits camera possibilities (forget about fact ProRes is lightly compressed and 4:2:2 etc)?

You gain more headroom in post with a 12bit output. If you going for greenscreen or rotoscoping than it would matter. Most monitors are not able to see all the colors from a 12/10 bit signal, but the information is there. Anyways… It’s more of a luxury problem IMO. A colorist will probably find 10bit 422 footage good enough in most cases as with 12bit RAW. I believe we are now talking more about numbers than real impact on what the camera can do or not.

Someone have added footage from a zoo.. recorded in RAW 1:4. I believe it would be just as good recorded in HQ10bit422 because of the grading. Maybe he was thinking about saving space rather than handling files.

https://www.youtube.com/watch?time_continue=1&v=u-Wy0CUV2EY

My question is: does 10bit recoding format really limits camera possibilities (forget about fact ProRes is lightly compressed and 4:2:2 etc)?

You gain more headroom in post with a 12bit output. If you going for greenscreen or rotoscoping than it would matter. Most monitors are not able to see all the colors from a 12/10 bit signal, but the information is there. Anyways… It’s more of a luxury problem IMO. A colorist will probably find 10bit 422 footage good enough in most cases as with 12bit RAW. I believe we are now talking more about numbers than real impact on what the camera can do or not.

Well, this is exactly the point- real data, not numbers. Those quoted number are so often exadurated
It's not me (but others) who said that ProResHQ 10bit will be a problem. I'm saying opposite. I'm talking about "this headroom" (not what you see), what you can pull off without footage breaking out.

Someone have added footage from a zoo.. recorded in RAW 1:4. I believe it would be just as good recorded in HQ10bit422 because of the grading. Maybe he was thinking about saving space rather than handling files.

https://www.youtube.com/watch?time_continue=1&v=u-Wy0CUV2EY

Very video alike. Don't like this one at all.

In my experience the difference between 10bit and 12bit (independent of codec)

Same DR for both
But a less malleable image in 10 bit.
- you capture the same DR however, if you have tones in the near clipping or near blacks then it becomes more difficult to reposition these in the grade, pull a soft key in the grade, grab a secondary.

DR ka the same, but image rubustness it quite different.

In the same way that an 8 bit image can “capture” the same DR but falls apart as soon as you try to grade it.

JB

I have to say, I was quite impressed with the dynamic range displayed there - the contrast range in some of those harsh daylight shots is pretty crazy, and the highlight rolloff seems really pleasing.

In my experience the difference between 10bit and 12bit (independent of codec)

Same DR for both
But a less malleable image in 10 bit.
- you capture the same DR however, if you have tones in the near clipping or near blacks then it becomes more difficult to reposition these in the grade, pull a soft key in the grade, grab a secondary.

DR ka the same, but image rubustness it quite different.

In the same way that an 8 bit image can “capture” the same DR but falls apart as soon as you try to grade it.

JB

Thanks John, useful info.

I have to say, I was quite impressed with the dynamic range displayed there - the contrast range in some of those harsh daylight shots is pretty crazy, and the highlight rolloff seems really pleasing.

Actually you are right- it just needs small Resolve touch to put those highlights even more down and stop them popping out. In the same time there is still something video-ish about it.

...and a grade of RAW footage. Added some motion blur as it is way to sharp to my taste.

Davinci_grade_15_ 11.18.47.jpg

Yes Bunk. Seen the footage, good. The few hdr glimpses, with sun, are good. But mid summer high day sun, in the tropics is what to try.

You can even lower debayering to 1/4 resolution and it doesn't change anything (GPU is only at 30% for full res debayering, so it explains it). Problem is CPU which is close to 100%. DNG is for whatever reason so slow.

That's odd... my experience has been that cDNG is quite fast in Resolve, as long as the disks were fast enough to keep up.

Take same clip as final video- ProResHQ etc. and even this plays way easier. This is not normal as DNG should be way easier to decode than any intermediate codec (RAW is just a monochrome image with 1 channel). It's like 3 or more times slower than it should be.

Not quite. De-Bayering is pretty expensive since that's where the interpolation algorithms run to calculate the colors that weren't captured initially. ProRes is already de-Bayered, so it's just a matter of loading and uncompressing -- it's designed to be computationally lightweight, so it will always have a lower CPU burden than any raw format ever will at the same resolution.

I'm talking about real usable data. I don't think that you will get much advantage from 12bit RAW over 10bit ProResHQ when it comes to headroom for adjustments. Point is that 10bit ProResHQ isn't really a problem as a recording format. It doesn't really limit what camera can actually record (maybe tiny bit).

You're right that it's not a problem as a recording format, I just find it pointlessly bulky. I stopped bothering with simultaneous proxies when my DIT realized that the proxies were as big as our raw clips... and the raw clips were 4x the resolution (8K vs 4K).

As far as raw data goes, it's definitely there. BMD has had that figured out since the BMCC, and has only gotten better overall (yes, I know some things went awry along the way, but for the most part, BMD has been improving with every new model).

The question isn't whether or not there's more real data available in raw vs ProRes, it's whether or not you need it for the project at hand. If you don't, then don't worry about it, and if you do, record it.

That's odd... my experience has been that cDNG is quite fast in Resolve, as long as the disks were fast enough to keep up.

Well- I assume that once you have fast 6-8 cores you don't notice it as this is enough for most 4K sources to play realtime.
RAW recording is the best for size and quality (also costs least processing in camera, so also saves battery). All what you need is nice solution for editing- Cineform is good example. Easy to play, simple but good enough debayering built in decoder, so RAW can be preview/converted if needed like "normal" video. Due to wavelet nature you have fractional resolution decoding (so 4K editing on 2 core laptop), so your single, main RAW files are at the end: full quality RAW data, "normal video" and proxy if needed. This is still so underrated solution. Instead of this we have plain format which decodes badly, workarounds with optimised/proxy media etc.
Many post solutions are so legacy and outdated. There is real lack of innovation in post industry- maybe because at the end it's not easy to make money as company which develops apps for post.

Guys, check out the new NVIDIA Turing based GPU cards, they will greatly speed up work flow when used properly. They even can do Hollywood raytracing 3D renders real time on the highest offering, which means even the cheapest can be used practically, just not real time, but real time for video work. So, it will be interesting when that hits down to a $1000 card. Our system prices can plumet as even an future iMac pro with one of these might handle 8k workflow with ease (and some future good disk options).

Yes Bunk. Seen the footage, good. The few hdr glimpses, with sun, are good. But mid summer high day sun, in the tropics is what to try.

Thanks for pushing! ...although still no tropics ;)

Davinci_grade_15_ 10.02.13.jpg (228.95 KiB) Viewed 3406 times

I live in the tropics, direct sun overhead. There is a lot more light than in many places people that buy these cameras come from. So, more contrast between bright and shadows. I remember the filming of MI2k, they wanted to show off the stunning light in Sydney, and that's a lot less than here. Anyway, the low horizon sun in the clip is heaps less light. I'm actually watching Episodes 7+ of series 6 of Longmire, and I wish I knew which camera they used from episode 7, it's so horrible, so low latitude, nothing compared to the dragon footage they used to have. So, it's important the camera doesn't start to look like that. If you are working with dark people and they are under cover mid day, it can be hard to do a reference shot from outside and not blow things out, and once under the dark tree etc, stop the outside greenery from blowing out while giving good justice to them. It's not racist, the doctor who delivered my father and discovered Ross river fever, used to have his clinic under a Mango Tree (this is traditional pioneering territory). It's what everybody used to do.

But then again, I think sun lit scenes are sexy, so like the latitude to handle strong lighting and sun backed shots. Some might like to use strong shadows (like in the episode 7 above which tried cinema style, looked a lot worse than the previous style), but with strong light instead, not leaving the shade get too dark. This is a special effect, not all the time.

You know; maybe I'll be the only one to say this, but "eh."
The camera looks fine, it really does, but beyond the pixel peeping, honestly, it looks like just about any other cameras "test footage" I've ever seen, in environments where you're really not pushing the thing at all.
It all looks relatively "pretty," though sharpness is a bit of an issue on skin (time to bring back a glimmer glass it seems), but it is also rather ordinary
The balloon stuff looks similar to what I've seen on a DJI promo for their phantom drones (and X cameras).
The models looks OK but it's really no different than what I'd've seen off of a GH5.

I don't mean really to diss the camera too much; but it would be nice, one of these days, if we really got some footage which shows how it could be used in a narrative environment. Something akin to what Arri did with the LF in the "papa" short.
https://www.arri.com/largeformat/

My 2 cents at least.

Eh, sure Adrian Most of the footage I don't regard much of a test. It's like a bit here, a bit there, a little of this or that. But great for testing grading

I'm more doing levels peaking. I dont think I have really started pixel peaking except on the contouring and shaping of highlights etc. Pixel peeping is fun to see all sorts of stuff on cheap cameras and in compression, which is not much of an issue here. Somebody complained it's too niosy, that's about it for pixel peeping. Maybe it is. I would imagine for most of us it's fine.

But as I pointed out, what about the different ProRes modes and settings in those modes?

BTW, at least some of those one inch drone sensors are Sony with similar performance. So, how come not pocket earlier.

Well- I assume that once you have fast 6-8 cores you don't notice it as this is enough for most 4K sources to play realtime.

Especially with a Pascal GPU... or a Turing.

All what you need is nice solution for editing- Cineform is good example. Easy to play, simple but good enough debayering built in decoder, so RAW can be preview/converted if needed like "normal" video.

I'm really hoping that Cineform gains some popularity now that it's been opened up, though explaining why would just be repeating what you're already writing.

Many post solutions are so legacy and outdated. There is real lack of innovation in post industry- maybe because at the end it's not easy to make money as company which develops apps for post.

There is indeed. It was only a few weeks ago that Jarred Land announced the update to the Red SDK that moves wavelet decoding onto the GPU, taking advantage of some of the new features in Turing.

I'm hoping that we'll see more of that sort of thing soon.

Those two lots. Oh, the GPU doesn't exactly fit our codec processing and can only process a chunk at a time on card. You would think they never have heard of real time embedded coding. ? Change the codec to fit the GPU assistance, and stream chunks in and out. That could have happened over 12 years ago.

Of course they have. Remember the Red Rocket?

The hold up was technology. 12 years ago you'd have had to dedicate a 500-node Beowulf cluster to match the computing power of even a mid-range Pascal GPU.

Somebody suggested a certain parallel processing for video board solution to them. That's why it cost souch money at first. It wasn't till more recently they finally switched to a gpu based solution. However, I suspect the previous solution IP is at the heart of the cameras since maybe late last decade. I used to track a number of low energy parrallel solutions, and there were a handful of solutions. However, putting aside the nonsense I was told, it became evident that older gpu's could assist a lot more in decoding if you put in the work. But simply changing the compression routine to suite processing solution further enhances this. I think cineform dud that for CPU processing. Even if you loose 10-20% compression performance you can gain mass produced hardware which could be 10x+ cheaper to use than FPGA, but FPGA is getting better. But the sort of routine could transfer between newer GPU's. But custom hardware can be many times smaller and more power efficient than GPU's (if you can afford to make it). Most of a gpu's circuitry is not not needed or wasted compared to a custom circuitry. So, while a GPU is handy for price at lower camera numbers, you could somplyale one of more sections custom to your codec, and maybe get similar performance from those sections compared to using the rest of the GPU. So, Red's issue was, that gpu's general program ability was not conducive to jpeg2k wavelet processing back then. Now we have much beefier gpu's, with whatever instruction set improvements, and faster more cores, things are different in recent years. David Newman discussed with me years back, their version of wavelets was customised for better communication processing performance. You could likewise do a cuustomise to maximise GPU processing assistance. You might say why bother, because your GPU cards can do it, but if you want a pocket camera to do it, you want to use the GPU in an arm based chipset to save on power, heat and costs. But, to split your processing pipeline back and forwards over multiple resources is likely a very sophisticated embedded deterministic parrellel computing routine compared to the CPU routine. A pain for even an experienced person in the area to program.

However, I have my own ocean of ultra low powered cores alternative to CPU and gpu's for my own raytracing technologies proposal on the back burner. I hope to archive cities less than 10k transistors or less (+memory) running at 5ghz+ (though it needs to be slower due to the packing density to stop it overheating) on conventional chip processes modified for some of my own circuit techniques. I was hoping to outdo conventional custom chip design on price and energy. This is far more performance per unit of area). But I'm also working on some next gen circuitry techniques to go past a tetahertz at fractions of power. I had come up with a number of proposals to work up to, but found that the most advanced proposal could work on a very old proposal I had. It's a shame my mind has shrunk so much, it all seems so blurry now. Like a bunch of blurry paths rather than a bunch of clear branching path networks of knowledge all balancing out. Now, I have to more and more, rely on trying to remember my previous working out. But there are so many different ways to do and enhance things.

Okay, okay, one more: What do you guys think?

Looking very good me thinks. Having a Nice looking model doesn’t hurt either.
But seriously, nice shadow and skin detail. Overall look is a little cool looking with the color grade going to blues.
Cheers

What do you guys think?

It really depends on how you like the grade, which (to my eyes) is too loggy and much too teal. I took out the teal and brought the blacks to zero, and even with the h.264 source, it looks good to me now. But that will always be a matter of taste.

Looking very good me thinks. Having a Nice looking model doesn’t hurt either.

Stuff it Jim (Denny) I'm only interested in models of cameras.

We have a near record drought over here, yet most of the samples BM has published has overcast days or less bright extremes of light. It's really starting to glow bright here at the moment.

Still seeing a lot of little issues in samples in respect of use for cinematography. I've got the bike footage on my TV last night, and Lattitude is stretching it a bit, and colour Lattitude issues, can see that in places here too. It's not a fail, the image/lighting just has to be worked more to get the creaminess of cinema. In the bike sample definition in the client uds through the trees was lost in order to expose for the main scenes. If you see those cheap 19+ stop hdr car cams, the sky is rich blue and the clouds have rich details and toning in them during high sun (like 4 years back). Seeing those clouds float by through a canopy, would really add 3D crispness to a scene. That's why 15+stops is important. The scene looks more naturally balanced and rich, and you can bring the clouds into focus to the scene without having to concentrate on them. It then looks beautiful unnaturally natural. Where is real life you have to refocus your eyes to concentrate on the detail of bright clouds, in a hdr scene like that the audience can glance the beauty a d architecture of the clouds without having to refocus and squint at them.