Bruno

My problem is not justifying grain, as I said I love it and I'm always inclined to use it. My problem is also not getting a well encoded h264 out of it.   My problem is that most of the online sources where people will watch my work are not controlled by me. Regardless of how good your h264 file is, it will be reencoded by youtube or vimeo, and I usually don't like what that does to the rest of the image, and in that sense, adding grain might be actually hurting the image people will end up seeing.   So what do you work for, the master you see at your workstation, or the compressed version everybody else will get to see? Every decision that degrades image quality or aesthetics is a hard one, but shouldn't one be made in this case?   Maybe toning down the grain could be a good compromise, maybe we need some further testing to find out where the issues get worse...     And of course BBC won't ditch all the 16mm content they shot throughout their history, but 16mm film is out of the question for any new content they produce.   “The BBC and most of the UK broadcasters now share a common delivery standard, and as all broadcasters are moving to HD capture and delivery - 16mm will not be acceptable, we have gone over it so many times since 2006, that unless there’s some real radical technology change it is not going to work. The grain compared to the neg size is too large.”   Ian Potts (BBC Executive Producer of HD)   Maybe h265 could improve things? But how long until it will be adapted?

This is not really a question, more of a survey to see what your experiences are in regards to adding film grain to digital footage.   I love film grain, there's a bunch of nice film grain packages out there, paid or free, that can add beautiful real 16mm or 35mm grain to your footage, and it's very easy applying it to any project.   This is all fine, BUT... most of the work I (and most of us) do these days is seen online, with H264 compression ranging from decent to bad, depending on the connection speeds and where it's found, etc. and H264's compression quality is way worse when using footage with grain than it is when using clean footage.   Using clean footage makes it much easier on the encoder to get better results out of the same bit rate, I could increase bit rates, but that's not the issue, since once it goes on youtube or vimeo, it's out of your control anyway.   So that's my dilemma, my Prores masters with real film grain look beautiful, but apart from a few select screenings or film festivals, no one is really watching those, they're watching crappy h264 versions online.   This is also the reason why BBC won't accept 16mm material for HD broadcasting, since it's based on h264 and the compressed 16mm grain wasn't deemed suitable for broadcasting.   What do you guys think?

  5D3 has two card slots, you can have ML on an SD card and shoot on CF cards.

  Yeah, just a few days ago Andrew was saying it would be highly unlikely that Canon's upcoming DSLR cameras would be capable of a 10 second raw burst, and now on the "old cameras" we have a 40 second limit when shooting continuous raw, and that's because of the file system, not the camera! :)   I've said this before, I'd rather pay for firmware upgrades that give us more functionality to an existing camera than buying new camera bodies every year. Mostly for environmental reasons, but this would also give Canon more time to work on the software side of things, instead of constantly developing new hardware, with small technological improvements, they could work on bringing the firmware to the next level.   This proves that could totally work, there's a lot to be milked out of a 3 or 4 years old camera if you have proper access to it, something that ML doesn't have. A lot of technology surrounding the camera evolves every year, such as card speeds, and features that weren't possible 2 or 3 years ago could now be implemented on that same hardware.

If you just want prores you could do that with any camera that outputs clean hdmi and some external prores recorder like the hyperdeck, but because of the way they've been processing those 'clean' hdmi signals, the quality is not the same as generating a prores file from a raw one. You can't do that at the moment with the 5D3 or any other DSLR I know of, would be great if they could redirect that raw feed to an external hard drive via usb or to a field recorder via hdmi, and seriously, at this point everything might be possible... In the Slumdog Millionaire example, they were using a SI2K camera, which is a small 2k raw camera brain, and it is meant to be used like that, or with any other field recording solution, it wasn't really a hack. Using ice on "burning" batteries under severe weather conditions is also more common than you'd think :)

Totally agree. Let's get one thing straight though, this is huge, this is amazing and the difference in quality is just unbelievable, and the way it's been going, I wouldn't be surprised if they sort out the 40 seconds limitation within the hour, but yeah, 40 seconds is a MASSIVE limitation for any kind of narrative, fiction or documentary, and anyone who claims otherwise can't know what they're talking about. You can find a million films without a single 40 second long shot, but probably none of those shots comes from takes shorter than a few minutes. Tell an actor they have 40 seconds to get it right and good luck getting a good performance out of them... I'm very excited at the possibilities this hack currently presents, I wouldn't be excited about using it as it stands. Some things haven't been answered yet either, I keep hearing that the camera is also doing less work since the raw feed is always happening anyway, but CF recording alone is enough to heat up the camera, and some people did say the camera heats up much more than usual when recording raw video, would anyone care to compare? Also, what about compressing that raw video to h264 in camera with decent h264 transcoding? Transcoding those raw files to h264 gives us much better results than the camera's own h264, so maybe they could override that and give us higher quality smaller files and longer recording times? Obviously Prores would be awesome, but that would probably not be possible, whereas with h264, perhaps the camera already has some sort of hardware optimization. Most of the footage I've seen also appears to stutter every now and then. Are we getting a steady 24fps or is it not accurate yet?

They already do it with standard h264 video files, even though it drops a few frames in between files.

Any 16mm lens you can mount on it!!! C-Mount, PL-Mount... Too many choices might be intimidating, Full Frame, S35, S16 and maybe S8 would be more than enough. Maybe they could make each of the crop sizes slightly smaller to accommodate for most lenses' coverage.

Also, a 16mm sized crop would be great. Full frame, 35mm and 16mm crop modes would give us the best of three worlds, leave no lens unused!

  These days, they could probably add more than twice the RAM for less money than what the 5D3's RAM cost at the time, so it's not even about that anymore.

Well I get the feeling they're stepping up their game a little bit, they postponed the 70D and latest rumors said it might come out with the 7D2 sensor instead of the SL1 one. I'm not saying I expect them to give us raw video though, I'm just saying that it's silly to be wishing for a 1 second raw burst. I understand it's quite cool to dig that feature out of an "old" camera, but that would be a pretty lame feature to support officially, or to be wishing for in a new camera, especially when you have a $1k 2k raw camera out there now. Kind of.

BMD are doing it for way more than 10 seconds at 4k with a camera that costs not much more, so obviously it wouldn't be impossible, would it? At this day and age I don't see why a 7D2 couldn't shoot 2k raw for constant video if that's what Canon decided to, not just 10 seconds, hell the BMD pocket camera does it for half the price, so why the skepticism? Asking for a 1 second raw burst, or even a 10 second raw burst, is thinking extremely small, if you're gonna wish, wish big, we don't want canon or anyone else thinking that's all we wish for, do we?

  The basic law for taking pictures in public spaces is "if you can see it, you can shoot it", she was being silly.

  Is there a precise definition for burst length? Could be a 1 second burst, could be a 10 second burst, that's a huge difference right there, still pretty limited though, for most types of video shoots, pretty cool for stills though.

  But we don't know for how long yet...

  You mentioned some of the reasons in the article yourself... this camera is still in a very general segment, and they would have a lot of tech suppoty to do once people started to shoot 3.5k DNG sequences and tried to open that in their home computers...   Also, "1 second video burst" doesn't sound like a very appealing feature for most of the consumers, on the contrary, they'd probably see it as a ridiculous feature (and you try to explain them why they should shoot huge raw files), because like it or not, there's probably more parents buying these cameras to shoot their babies than indie filmmakers.   Then there's the reliability issues, overheating, etc... This is far from being a feature, it's a curiosity, and you either make it a full feature or you keep it out. While even a 10 seconds burst would feel short, maybe they could market that somehow, but not 1 second.   It's definitely cool to see the potential, but that's all it is so far. I'd love to see it's possible for the camera to shoot constant 2k raw video though...

Maybe they could even try sending a better quality (raw?) stream through HDMI?

  Yeah, I wondered a few times if it would possible for them to just send the data through USB to a fast hard drive. Not sure if hard drives would be faster than modern memory cards though.

Can ML hack the CinemaDNG file creation, or are they just copying it as it is? Can they access compression settings and make those files smaller?

  So you could get a 4x bigger buffer and record 4 seconds of video? :)

I know about Apertus, but it just kind of proves my point, their project and feature set looks less and less groundbreaking every day, and things are moving very slowly, compared to the rest of the industry.   The Digital Bolex backers will probably get much more than they expected, since there's been over 100 significant improvements to the camera since it went on kickstarter...

As far as we know they never built a single camera, all their experience is with software/firmware, so would that really be a good idea?

Does it capture interlaced footage when using 50i or does it just output interlaced footage from a progressive captured image? (like the Sony RX100 for instance) In that case it would be pretty simple and effective to deinterlace it.

    You know an LCD screen is not the same as looking through a viewfinder... Also, they've been using measuring tape to focus for over 100 years of film, maybe more than their eyes.  :)

Nice review Andrew. I get te feeling that whenever you mention image quality you're referring to resolution and sharpness. While those are definitely important factors, there's other factors like color that are extremely important, and for instance when you look at all those differen camera frames, Canon's image is by far the most pleasing, regardless of the resolution, all others look colder and more video like in comparison. Of course a lot can be done in post, so it would be nice to see how well it grades, which profiles look the best, etc.