Does Canon ever talk to their customers?

[Barry Goyette]

Originally Posted by Jon Fairhurst

On paper the C300 looks better for fast turnaround and the Scarlet-X looks better for heavy grading.

I think this is a big question for me. My 5dmark11 files seem to fall apart when you try to push beyond a little color or tone adjustment. And having never worked with a red file I'm certainly no expert but given the following assumptions: an extremely clean, properly processed 8 fit file (c300) converted to 10bit for grading versus a flat raw file (scarlet-X) that is perhaps not as clean a capture...is it an absolute fact that the red is more gradeable? (and that was a question not a statement).

At the canon event, we definitely saw films that weren't heavily graded. I think the talking points that came out of the viewings really were the hi ISO, dynamic range--highlight and shadow detail, "film noise", and from xxit, that the camera can work in virtual studio/effects situations, in addition to the primary message of a light, inexpensive production tool. The one film that seemed to have a "look" applied was "Sword" and I'm not sure it was that successful (to my eyes it seemed to be an attempt to squash some smoky shadows due to high ISO filming).

I'm not sure canon was attempting to shove "grading" behind the curtain...I think the biggest message was simply that you can use this camera for theatrical distribution thus striking 35mm prints, showing the films on a big screen and showing looks that had flattering skintones, and thus a look that was pretty mainstream. But it's still a question in my mind. I think Vincent Laforet was trying to get some ungraded footage up last week, but I don't know If he was successful.

[Jon Fairhurst]

Originally Posted by Chris Hurd

Since it's recording the Canon XF codec, the list of approved CF cards is exactly the same as the one that appears on the Canon XF 305 / 300 product pages... absolutely no difference in that regard.

This is great info. This place rocks! :)

[Jacques Mersereau]

Originally Posted by Barry Goyette

I think this is a big question for me. My 5dmark11 files seem to fall apart when you try to push beyond a little color or tone adjustment. And having never worked with a red file I'm certainly no expert but given the following assumptions: an extremely clean, properly processed 8 fit file (c300) converted to 10bit for grading versus a flat raw file (scarlet-X) that is perhaps not as clean a capture...is it an absolute fact that the red is more gradeable? (and that was a question not a statement).

Horses for courses.
What depends is how hard one is going to push the grade. 8 bit @ 50mbps is fine for a lot of stuff, but
raw@100-440mbps is going to give a ton more capability.

Convert 8 bit to 10 bit? I am not sure if adding two zeros does little more than increase file size, thought it might help slightly if processing the data through a few differing crunches. Others can weigh in with their experience.

As far as nothing but negative comments, please re-read my posts. I said the C300 made beautiful images and most purchasers would be very happy. BEAUTIFUL IMAGES AND VERY HAPPY.

"Negative comments" are not mine alone and not personal - like yours Sir. I am far from the only one who sees some gaps in what might be SUPER AWESOME. Please remember, it was Canon's vast over-reaching marketing spin that does NOT meet the camera reality. That kind of chest thumping is not savvy or necessary unless you can back it up in spades.

[Chris Hurd]

I don't understand what you mean by "Canon's vast over-reaching marketing spin." Canon has not yet begun to market this camera. And it's going to be a fairly hard-to-get item, considering that only a handful of dealers can even sell it. It's not going to be seen all that much outside of broadcast television production.

The Hollywood event was more about the Cinema EOS line (keep in mind that all but two of the products in that line are lenses) and their new Hollywood office and lens repair facility.

[Jacques Mersereau]

I looked around for the original Canon announcement - maybe you can post it for us. I might be WAY wrong here, but I seem to remember something about the 'event' being some kind of industry changing move by Canon - something that would alter Hollywood or some such. The anticipation built on various boards with Jannard going 'all in' by announcing Scarlet would also launch Nov 3rd.

Whatever it was Canon planned, Canon would not even confirm it was a camera - so the rumors spun up the heat (and maybe that was the hope?) that 'this game changer' could be anything, but it was certainly bound to be incredible.

Then, pip, the 1080P@30 fps, 8 bit, $20K 'industry rocker' near fizzle. SORRY NO SONIC BOOM HERE.
The _only_ reason it isn't a total fizzle imho, is because the images look so great and the C300's super low noise - low light capability. And sorry, I will say it again, PRICE MATTERS. Had this thing come out under $10K it would rock. The idea of a 5DMII with mic inputs and SDI out for an additional $16K just leaves me feeling like I was kicked in the teeth. Others feel differently, and that is cool, but so far we have ONE taker and even he is now asking about GRADING. Ask yourself and be honest, is the C300 the camera you would recommend for $20K if the buyer said he wanted the ability to really push the grade hard? And please, we know the answer.

[Jon Fairhurst]

That topic (Nov 3rd announcement) has been discussed at length. The bottom line is that it was an historic announcement from Canon's perspective (new market, products, sensor, business strategy, support center.) It certainly wasn't an historic announcement from the perspective of typical Canon DSLR shooters. People who expected a 5D3 or AF100 competitor were disappointed.

But that horse has already been beaten into submission.

[Allan Black]

Didn't I read where many cams from the first batch will be offered to US rental outfits? Makes good sense on Canons part.

Cheers.

[David Heath]

Originally Posted by Jacques Mersereau

Please remember, it was Canon's vast over-reaching marketing spin that does NOT meet the camera reality. That kind of chest thumping is not savvy or necessary unless you can back it up in spades.

Jacques, all I remember directly from Canon is one single fairly anonymous "poster" advertising a "historic" announcement on November 3rd, which I first saw when Chris posted it here. ( http://www.dvinfo.net/forum/new-cano...ml#post1682321 ) I don't recall Canon themselves actually saying very much at all after that, certainly not "vast over-reaching marketing spin".

What did happen was that Red made their parallel declaration about also making an announcement on Nov 3rd, and Red did have quite a lot to say in the run up. Let alone all the speculation on forums such as this. But virtually nothing from Canon themselves - I certainly wasn't aware of "chest thumping" on their part. (And I have no direct or indirect vested interest in Canon or any other manufacturer.)

[Dermot Shane]

Originally Posted by Barry Goyette

an extremely clean, properly processed 8 fit file (c300) converted to 10bit for grading versus a flat raw file (scarlet-X) that is perhaps not as clean a capture...is it an absolute fact that the red is more gradeable? (and that was a question not a statement).

Speaking as someone who grades films.. yes, absoutly r3d RAW has more options

And there is no such thing as "converted to 10 bit", 0-255 is still 0-255 even when inside a 0-1023 space.

Think like.. i buy 256 ounces of a good wine, and i pour the bottles into a 1024 ounce bucket... i don't somehow convert the 256 ounces of wine into 1024 do i?

I just have a half full bucket.. and a pile of banding issues on smooth graduated surfaces.. like walls, and the cheeks of softly lit leading ladies...

d

[Barry Goyette]

Originally Posted by Dermot Shane

yes, absoutly r3d RAW has more options

I totally agree. With R3D files, you can definitely go more directions than any 8 bit file. (I stated exactly the same words as you on another thread last week). I think the question that I meant to ask (and this is totally hypothetical--since there are no canon mxf files out there but one ultra-high iso sneaked shot from the event), is could using a combination of in-camera adjustments and "significant" post grading on the canon produce an equal or better file (1080P of course) when compared to scarlet. Or conversely, is it a given that an R3D will always produce a superior graded image to the canon mxf.

For the sake of argument, lets put this in "highly favorable for canon" framework. Two identical clips shot at say ISO 6400 or higher. One on Red 4k, the other on C300. From "what I've read" the Canon should produce a virtually noiseless image with a full 12+ stops of DR. Let's say that the DP was good and was able to get close in camera on the Canon using the internal 12bits to produce a great 8 bit file. From there, significant but not severe grading ensues. Now to the Red. From what I understand Scarlet would carry a fair amount of noise at this ISO and as Scarlet isn't capable of HDR, would probably be out of bounds on it's optimum DR. Now would that "flat, raw file" be able to produce as good a file as one off the C300?

Sure it's not a fair comparison...so how about at 3200? 1600? 850? Where's the line?

As I understand (and I may be wrong) the red "raw file" is true raw, so other than exposure parameters, all other adjustments are applied as post processing of the raw file. In the canon workflow, some adjustments can be made internally in 12bits prior to the production of the 8bit output file leaving less to do in final grading. So back to the top, red is definitely more flexible... but is it necessarily always more gradeable? (and it's ok to say yes here.)

And there is no such thing as "converted to 10 bit", 0-255 is still 0-255 even when inside a 0-1023 space.

Mostly you are right about this, and I'll not pretend to be an expert in video grading. However, I am an expert at grading in photoshop (as I shoot on a hasselblad, I have a pure 16bit workflow) and some of the techniques I use when bringing in outside 8bit images do see some benefit from bit upconversion. When I wrote the question last week, I did a little comparison here using some standard layered grading in photoshop, using an extremely clean 8 bit image and processing it identically in both 8bit and 16bit workflows. There were differences: both in the histograms and in increased visible banding in the 8bit workflow. It was a subtle difference, but one that did in fact matter. My guess is there are similar subtle differences obtainable in video as well. Furthermore the sensor does play a factor in this part of the equation. I know (again from photoshop) that 16bit files are not all created equal. A properly processed Hasselblad FFF, output to 8bit file is nearly as adjustable as a 16bit tiff output from a 5dmarkii CR2 because simply put, the 5d sensor isn't as good, thus the raw file isn't as clean, and therefore it's 16bit output is really something less than advertised.

Again my original question was purely for discussion's sake. As has been said here now many times: Both these cameras are most likely quite capable of outstanding images in the right hands.

[Dermot Shane]

Hi Barry..
today finds me juggleing three features, two of them were both shooting exteriors in the same city at the exact same day & time and i have both shows on my server currently.... the files have time/date stamps

in the one i''m just doing a final pass on currently there is one scene that they had lost sunlight, and shot anyway.. and the REDmx was clearly underexposed when the r3d's were debayered with metadata, but once i changed to print density and ran the ISO & noise reduction up the shots were recoverable, actualy it's a very nice scene... if you watch the film in your local cinema you would not know.

the feature i'm starting to grade next week is shot on F3/444/Slog.. also shooting at the same time, just down the road from the first crew.. the F3's footage looks very clean indeed when shooting into magic hour.. but it does lack the flexibality of the raw workflow.. i can't really make it be anything other than what it is already without jumping through hoops

Now all of these are being finished in Dci2K / P3 so i'm looking a colours that can only be reproduced by a digital cinema projector (or a very small handful of OLED & LED screens), and i'm looking at it on a 16 foot screen.

I don't think any YUV/422 50mbs 8bit codec would stand a even the slimmest of chances in this enviroment (2K/444/P3) - no matter how good the camera is, actualy any YUV/422/8bit even uncompressed would be challenged i would think

Most everything i have worked on has been shot on RED, film, or Alexa in the last few years, but i did grade a doco feature last spring shot mainly on EX3's - so 8bit/35mbs/Long GOP codec - a very good codec... it was far from easy... really, secondaries were tearing like crazy - it was so very limiting... basicly anything beyond a first pass is not possiable... i would expect much the same with the C300.


d

[Barry Goyette]

Great post Dermot. Thanks for the short look into the complexity of what colorists like yourself are up against daily. It makes your perspective on the cameras much clearer.

[Brian Drysdale]

I think how much manipulation you want to do in post really depends on what you're trying to achieve.

Having put looks directly into broadcast cameras with comprehensive paint boxes, you can do quite a lot using that method, including the bleach by pass and other "looks". Of course, once set, you can't change your mind about a look strategy, which can be more a producer issue than a DP one or perhaps even one for the director.

On the other hand, DPs do put their look into the LUT for on set viewing when shooting RAW and the final film looks nothing like they were trying to achieve because everything has been changed during the grade.

From what I read by an extremely fussy DP who had a quick grade using the rushes from the one shorts, the images don't fall apart when pushed. He mightn't have been doing as much manipulation as been suggested here, but he's positive about what he's seen to date (regardless of the numbers) and thinks the C300 should considered as an option for low budget TV drama series.

Canon seem to have other cameras in the pipe line, so it appears that the needs of people needing more post work will be met by those cameras rather than this one. Not all productions have a lot of grading done in post and the C300 seems to be aimed at them.

[Alister Chapman]

When your talking about pro level camera, anything from an EX1 upwards, IMHO the first thing that will make it hard to grade an image is noise. This is simple enough to confirm, take two shots with the same camera at a low gain/ISO setting and a high gain/ISO setting but exposed so the recording levels match, then try to grade those two images. The difference is normally striking.

I hear a lot of talk about shooting at very high ISO's and really have to ask why? I have an F3, it's 800 ISO when using S-Log, frankly I'd rather it was 400, not because it's too noisy, but just because it's really more sensitive than I'd like. There's lots of talk of using high ISO to reduce the amount of lighting that you need, but in reality that argument doesn't work. If you shooting a lit set, maybe with practicals, then you'll still need decent lamps to light the seen so the practicals are not blowing out. If your shooting outside during the day then any fill lamps still need to be just a bright as you generally working against the sun or ambient light. At night in a city your going to be working against the general street lighting, store fronts etc.

So what's key is how much noise the camera produces in the 400 to 1000 ISO range IMHO. This, I believe is the comfort zone we should be working in. More than this and the camera is becoming too sensitive.

After noise comes bit depth. Bit depth becomes more and more critical as the dynamic range of the camera output increases. An 8 stop camera using 8 bits is 1 bit per stop, not so bad perhaps, but when you start going up to 10, 12 even 14 stops then 8 bit struggles, there just isn't enough data. Now you can take an 8 bit output and convert it to 10 bit using a decent interpolation algorithm that will calculate the expected in between values. This can bring a reasonable improvement over raw 8 bit as it's not that hard to estimate what the in between values should be with some accuracy. It's not going to be as good as an original 10 bit source, but for not so severe grading it can work well.

Where you really get in to trouble is when you start using highly compressed 8 bit codecs with high dynamic ranges and a lot of noise.

All of this brings us back to the key differentiator between all these new super 35mm cameras. That differentiator is the operator. A skilled operator that appreciates the limitations of the camera and workflow that he has chosen will be able to produce a great image. I'm quite sure you will be able to grade C300 footage, provided it has been shot in a manner sympathetic to the way it's recorded and processed in post. Even with just an 8 bit output it will surely benefit from recording on to a higher bit rate codec, maybe with a NanoFlash or Samurai.

As for Red and "Raw" well if a lossy 5:1 compressed codec counts as "Raw" then so be it. It's only really "Raw" in so much as it has not been de-bayered. Which is more "Raw" non de-bayered but compressed or de-bayered but uncompressed (F3)?

[Don Miller]

Originally Posted by Barry Goyette

Great post Dermot. Thanks for the short look into the complexity of what colorists like yourself are up against daily. It makes your perspective on the cameras much clearer.

Yes, that is interesting. Especially considering it can now be done at the Scarlet price.

As far as grading and noise, the common use of the simple term "gain" by manufacturers is unfortunate as camera fall well out of spec at the highest settings. I think using ISO terminology when the camera is still in spec, and then Hi1, Hi2 etc. beyond the highest ISO is better as it makes it clearer that the resulting file might not match well.

A high sensitivity camera isn't just capable of shooting in dimmer light, it's capable of more accurately recording at all gain settings. The Epic likely has more read noise at base ISO than the F3/C300. That likely lowers true resolution.