Perya (Town Fair) - a Canon 7D Short
- Thread starter Jean Dantes
- Start date
Michael Olsen
Veteran
YouTube makes it look like there is a ton of skew in that, at least on my PC. I know there isn't as I re-watched the raw footage just yesterday...but that clip definitely shows it.
Taylor Rudd
Veteran
You've probably already done this, but you might want to shoot Philip an email asking/telling that you uploaded one of his clips
Taylor Rudd
Veteran
Just tweet him or something... http://twitter.com/philipbloom
I've always been wary of crushing the highlights. Obviously there are blown highlights from the kicker on his nose and forehead and...well...once its gone its gone with digital! If the information had been retained I think it would look good, but since its lost I wouldn't try to bring it down that much. Otherwise I like your direction with the CC. Once we get to fiddle with presets and have footage that is a bit more controlled*, I can't wait to see what you do with it.
*this is no bash on the footage that Philip provided. In the original clip the blown highlights look fine. It's just bringing them down that calls them out and creates that "muted" look
I've always been wary of crushing the highlights. Obviously there are blown highlights from the kicker on his nose and forehead and...well...once its gone its gone with digital! If the information had been retained I think it would look good, but since its lost I wouldn't try to bring it down that much. Otherwise I like your direction with the CC. Once we get to fiddle with presets and have footage that is a bit more controlled*, I can't wait to see what you do with it.
*this is no bash on the footage that Philip provided. In the original clip the blown highlights look fine. It's just bringing them down that calls them out and creates that "muted" look
Adam J McKay
Veteran
Really great.
Eddy Robinson
Well-known member
Just in passing, I was reading yesterday about new research at the University of California where they've developed a form of working lab storage (ie it only stores a few bits now, and will need at least 5 years to scale up) by storing an iron particle in a carbon nanotube. This should yield densities of up to 1 terabit/square inch (> hard disk) and thermodynamically stable for a billion years. Yes, billion.
They have built a working cell which can be written to and (non-destructively) read from with simple electric terminals using ordinary voltages (~0.5-3v, nothing unusual from an engineering standpoint). While work is needed to scale it up for high-density arrays for commercial use, there's nothing fundamentally difficult about that, and the methods aren't so far away from those already in use for silicon-based memory. So, you can figure on effectively eternal data storage by the end of the next decade, at worst.
Paper is here, if you like physics - it's quite readable, and a very elegant piece of work: http://www.physics.berkeley.edu/research/zettl/pdf/361.NanoLet.9-Begtrup.pdf
PaPa
Major Contributor
oh, i totally agree, i just wanted to get a simulation of what it could look like with a more controlled image, to see the contrast ratio.
Eddy Robinson
Well-known member
Respectfully, just as you think we don't get what cinematography is, I don't think you fully understand digital, if you describe a PC as a 'medium'. The storage medium is not the computer, it could be anything; on hard disks it's magnetic fields, on optical disks it's little scorch marks created by a laser, just above I describe a new technology involving lumps of iron in tiny tubes.
What makes something digital is the data structure employed for the storage of information, regardless of medium. As long as there is sufficient bit depth (as in Cineon for example), any analog information can be stored digitally with arbitrary levels of precision. I think we agree that digital systems can record any information you throw at them with arbitrarily high degrees of fidelity: our video cameras are limited in the dynamic range they can handle today, but honestly I think we'll probably be finding 4'4'4'-capable camcorders for free in our breakfast cereal within a decade - oK not quite free, but a sub $100 soundcard you buy today would have been world-class a decade ago.
As others have said, if you want to see the future of digital video, look at the past of digital audio. Now, you're saying that because film is organic - what you actually mean is analog, as there are no organic chemical compounds in film, as far as I'm aware - the behavior of the chemical molecules in the film stock when exposed to light or various washes in a developing bath. Now the behavior of molecules is non-linear, and that can be modeled - just as the behavior of current in an analog electric circuit is nonlinear, and has been modeled.
The only limitation is computer power. Our ears hear at much higher frequencies than our eyes see: the amount of information in 1/24 of an audio second, ie 1 frame of film, is about 4 kilopixels (I'm simplifying a bit here, mind). We use parallel techniques in video processing to handle the vast spatial resolution of images, whereas (mono) audio is one dimensional, but we have long ago solved the problem of accurately generating believable non-linear behavior in audio synthesis - indeed, the competition between manufacturers of creative audio software these days is not in super efficiency but in selecting algorithms and coefficients that sound 'interesting' in the same way that the imperfections of analog circuitry on older synthesizers, compressors etc. sound 'interesting'.
If we want to make video look like film at the prosumer level, eg in After Effects, we currently use what are called transfer functions, and most of the time we employ pixel sampling (again, I'm simplifying here, even AE often does these things in a more advanced way). The crappiest way of giving some video a film look - assuming we are already happy with the color - would be to take a scanned image of a film frame showing a flat gray and add the grain pattern to the video in proportion ot the brightness of a given pixel. You've seen this effect, it looks fake.
Better would be to use samples of a film frame showing a gradient, or a color card so we could do it by channel, or (better again) a sequence of film frames showing a color card under increasing light levels from darkness to blowing out. This is more-or-less what look-up tables (LUTs) do: so you can get close to the look of, say, technicolor film by referencing a big table of pixel values for levels of luma and chroma which are characteristic of a given film stock. This is still a single-step transfer function operating on a frame-by-frame basis.
That will get you most of the way there, to the extent that Joe or Jane Average won't know the difference at all, but an experienced eye like yours or (I hope) mine will notice it's a little too perfect and consistent and the light doesn't 'interact' with the digital medium as it does on film - for example, the grain pattern won't do interesting things around the border between two colors or in different parts of a contrasty image. Ironically enough, film manufacturers are constantly trying to reduce these inconsistencies on very high end stocks so as to offer film that offers super predictable and accurate image acquisition (offering the greatest freedom for manipulation later) while as artists we're fascinated by the inconsistencies that give film its exciting texture just as canvas and brushes offer a more expressive medium for painting than smooth paper and an airbrush.
So currently what happens most of the time is we take our 'boring' video image which is accurate but lacks personality, and add samples of personality collected from scans of film stock. By using LUTs we can add that in proportion to the content of our video image which gives us a much more dynamic emulation of film. This is akin to how a digital audio sampler works - make umpteen recordings of a piano key or guitar string played in different ways, and mix and blend those recording sin response to the playing information from a digital keyboard, like how fast and hard the keys are hit.
But where this is going is towards what's called physical modeling in the audio world: instead of recording a real instrument, you build a model of it (whether that's electronic circuits, or strings and a soundboard, or a metal tube with air blown through it) and synthesize the output by performing simulations of the physical system at very high speed. This was what they tried to do in the early days of synthesizer, but analog electronics weren't up to it and so we unintentionally wound up with an entirely new class of musical instrument which didn't sound very like what it was supposed to be emulating. Then samplers came along and everyone preferred those because they gave much more believable results. Then People bought cheap discarded analog synthesizers and decided they sounded interesting in their own right (which is how we got techno music). Then we started sampling those, but it showed up the limitations of samplers; then we started modeling analog instruments instead of sampling them, which techniques turned out to be equally applicable to acoustics, and round and round it goes.
What's happening now in high-end digital suites, working its way into prosumer software, and may eventually be built into cameras (because it's all software) is that we will synthesize the behavior of film stock by using iterative transfer functions. Future cameras (say by 2020) might externally say they're recording at 24 fps and a 1/48 shutter - interlaced vs progressive will be long gone by then - but in fact they'll sampling the image many times over for each frame, and for each sample they'll store or calculate (depending on whether you want to do it in post or in-camera) how that overly-perfect sampled image would affect a given film stock, then take the results of that and mix it with the next sample, and repeat the calculation.
Bottom line (finally!) is that rather than taking a snapshot of a moment in time and matching that up with a static description of a given film stock on a pixel-for-pixel level, we'll model the behavior of a piece film stock over time as an interconnected dynamic system. That probably sounds like something for which you'd need a computer the size of a house, but there's all kinds of wonderful mathematical shortcuts you can employ. The best analogy I can think of is that it's like simulating the behavior of a spring mattress as a weight falls on it - you have all these interacting springs which is what makes it bouncy compared to foam mattress, but the mathematics of simulating a spring are actually very simple and it's just a matter of doing a lot of calculations in parallel.
Of course, this sounds like a ridiculous amount of effort to go to just to deliver convincing emulations of film grain, why not just buy film instead of laboring away for a decade with computers? But the techniques described here are useful for everything from computer animation to simulating the behavior of fires, weather and so on. It's the same driving force behind the exponential gorth in power of graphics cards and the trend towards multi-core CPUs.
As a good guide, look at the quality of CG simulations of smoke and fire today. You can do stuff in After Effects now that stands up against the effects in The Matrix 10 years ago, which was the cutting edge VFX movie at that time. On today's cutting edge films, I really have difficulty seeing where the compositing is any more, and I have a pretty good eye for this and understand the techniques reasonably well. Simulating the behavior, rather than just overlaying the appearance, of film grain is well within reach right now, it's just computationally expensive to do so. But that expense falls by half about every 18 months, so what needs a big budget now will cost about 0.5% as much by 2020.
ydgmdlu
Veteran
Thank you for explaining, in obscene technical detail, what my layman mind was trying to articulate to another layman mind. My mind is blown; I am beyond impressed. I hope that you don't take that as sarcasm, BTW. :beer:
Eddy Robinson
Well-known member
...and this is how we end up with individually beautiful shots that don't cut together well :beer:
Look, I know you don't want to have to become a programmer in order to keep working as a cinematographer. But you probably don't want to be a chemist either, do you? If it was just about chemicals, budding cinematographers would try to get jobs at the Kodak or Fuji factories and spend a year mixing up emulsion and painting them onto plastic sheets.
In fact, you buy giant rolls of emsulsified ribbon which has characterstic behaviors when exposed to light and then to certain kinds of chemicals, and you know enough about these characteristics to look at a scene and imagine what you'd get if you took a particular stock and did a bleach bypass or pushed it a little or whatever. The only difference between this and digital cinematography is that you have to start thinking in terms of color spaces and articulating the characteristics the define the look you're after. Yes, it's hands-off in one sense, but on the other hand you won't need a fume hood.
Michael Olsen
Veteran
Hi Eddy.
This little bit made me wonder - is there anything organic in film stock? I looked it up and learned a lot!
Turns out...yes. The coating of film is commonly made of plastic - either polyester, cellulose acetate, or nitrocellulose. All of these are organic materials, as is gelatin, which bonds the emulsion together. My understanding is that the film grain, however, is created by the rather random arrangement of the silver halide particles - a non-organic compound. So it seems that film contains organic materials, but that these materials do not contribute significantly to the "look" or grain of a film...that's all in the silver compound.
Knowing virtually nothing about film, however, I don't know how much of this is right. If there is an expert around, please point out any errors.
Eddy Robinson
Well-known member
Oh come now, that's just ignorant. A file format is just a description of how the data is organised. Maintaining the ability to read deprecated file formats is trivially easy. Give me a piece of 20 year old software that ran on a 286 and it'll run just fine on the quad beast I use today. As long as the file descriptor has been written down somewhere (and programmers are compulsive backer-uppers) you'll be able to read it. Not only can I open an old GIF89 image in photoshop, I could write a program to do so in an afternoon because the spec is online. In fact I even have a few photos in that very format.
Old software doesn't die, any more than old mathematics does. Particularly not file format specifications that have made any kind of significant impact.
> Our ears hear at much higher frequencies than our eyes see
Audible sound is roughly 20Hz-20kHz, while visible light is about 400–800 THz.
> analog electronics weren't up to it and so we unintentionally
> wound up with an entirely new class of musical instrument
> which didn't sound very like what it was supposed to be emulating.
> Then samplers came along and everyone preferred those because
> they gave much more believable results.
I actually quite like the sound of the good old Moog.
Audible sound is roughly 20Hz-20kHz, while visible light is about 400–800 THz.
> analog electronics weren't up to it and so we unintentionally
> wound up with an entirely new class of musical instrument
> which didn't sound very like what it was supposed to be emulating.
> Then samplers came along and everyone preferred those because
> they gave much more believable results.
I actually quite like the sound of the good old Moog.
Eddy Robinson
Well-known member
Quote true - I should have said 'in the emulsion' - many plastics are made up of organic molecules. As you say, the interesting stuff doesn't take place in the substrate, though.
Rakesh Jacob
Veteran
IS IT ORGANIC??? This is annoying... because as some have stated it's not about organic it's about analog vs digital.
Organic: relating or belonging to the class of chemical compounds having a carbon basis
Your MiniDV cam has carbon atoms all through it, your HD camera, your hardrives, your PCs. Not just film stock. FOR GOD's SAKE WE LIVE ON THE PLANET EARTH!!! Carbon atoms are unavoidable!
Organic: relating or belonging to the class of chemical compounds having a carbon basis
Your MiniDV cam has carbon atoms all through it, your HD camera, your hardrives, your PCs. Not just film stock. FOR GOD's SAKE WE LIVE ON THE PLANET EARTH!!! Carbon atoms are unavoidable!
Taylor Rudd
Veteran
Semantics are awesome! *high five*