Digital Intermediate

[Brett B]

I just wanted to know what the disadvantages of using a digital intermediate are if the project was shot on 35mm. I know the benefits but am not sure what the downside is. Thanks.

[Daniel Henriquez Ilic]

Hiya,

I would say that if one have shot a long-feature (or so) in 35mm and one plan to post-produce it through digital intermediate for final distribution in 35mm prints, but that the resolution you will use for the digital intermediate is only 2k, you are putting in the trash a lot of details that have been captured in the OCN. If one could afford 4k or 4k+ digital intermediate for 35mm OCN, everything is ok. (Anyway the "near" future is 4k DI for 35mm, of course).

 

So I would say that in a generic way it is wiser to shoot on Super 16mm (with the new stocks) if you are going to do a 2k DI.

 

Regards,

Daniel

[Max Jacoby]

Besides the lack of resolution compared to contact printed 35mm, I find that DIs lack good blacks, even those printed onto Vision Premier. Personally I don't like the look of DIs because of what it does to colors and skintones in particular. While cold colors look fine, warmer colors tend to look fake. In particular people's faces become plasticy looking. I presume that is because the colordephth and resolution is not high enough yet. Since people's faces is something we are used to seeing every day that is were the drawbacks of DIs are most visible.

 

What will also greatly influence the look of a DI is the equipment and postproduction path that is used. The Spirit Datacine, although very popular, is not a very good machine to scan your neg. It is not even true 2K actually. Pin registered scanners like the Northlight and the Kodak Cineon will give you much better result. Then there is also the ArriScan (LED based) which is brandnew, but of which I haven''t seen any scans yet.

 

I don't agree with the statment that for 2K Super16mm is good enough. 35mm is dramatically sharper that Super16mm.

 

For the grading itself there seems to be a trend in American DIs to use grain reduction a lot, which will make the picture very smeary looking. 'Kill Bill' and 'Big Fish' are some of the worst examples in that regard. Somehow the Europen DIs I have seen (mainly French and English) tend not to do this as much. There grain is not seen as something one should try to get rid off at all costs.

 

On the outputting end, personally I have found the look of CRT recorders to be much nicer and more organic, with better blacks than the ever prevalent Arrilaser.

 

For an honest account of DIs I suggest you read Bill Pope's account of Spiderman 2 in AC. He is one of the few people who give you both advantages and drawbacks of DIs.

 

You should realize that a topnotch DI will cost between 250.000 and 300.000 Dollars. They can be done for cheaper, but the quality won't be as good. So far the best DI I have seen was for 'Troy', which was handled by the Motion Picture Company here in London. The French lab Eclair also seems to produce good results. On the other end of the spectrum there is EFilm who do a lot of films, but who have failed to impress me so far. They seem to be in love with DNR and the results can be quite appalling at times.

[Guest Daniel J. Ashley-Smith]

So I would say that in a generic way it is wiser to shoot on Super 16mm (with the new stocks) if you are going to do a 2k DI.

Most films use a 2k DI, why doesn't everyone shoot 16mm?

[Daniel Henriquez Ilic]

Hello,

 

Here we have seen in the cinema theatre (on 35mm print), the longfeature "Machuca" (2004) shot on super 16mm, scanned through Spirit(1) and with a 2k DI post-production (conformed on Scratch), color-corrected at 2k data res. quality and then recorded back to 35mm color neg. (through Nitro Extreme).

Really excellent resolution. The skin tones were ok. Beautiful image work from the DP and Chilefilms post-production team.

For me a very good example of very high quality DI work is The Lords of the Rings.

 

If last generation super 8mm stocks (Fujichrome Velvia, Eastman Vision2...) through Zeiss optics (or at least other optics with that quality range) results "equivalent" to a 2k data file (with 10 bit log bit depth), then, of course super 16mm digital "equivalent" would be around 3k. When I say "equivalent" I mean, that one thing is the objective resolution and color depth that could give the stock+the lens (in optimal photographic condition) but then as it is digitized, the sample need to be done at a highier resolution than the theorical resolution of the original film media in order to avoid A/D artefacts.

 

In that sense, in the sound post production world, the sample frequency is now around 192 kHz, with most digital in-field recorder working around 96 kHz, so that Shannon-Nyquist theorem is well applied. So that the 48 kHz and 44.1 kHz appears as a very basic application of the theorem.

 

Regards,

Daniel Henriquez Ilic

Edited November 5, 2004 by Daniel Henriquez Ilic

[David Mullen ASC]

Most theatrical films still don't use a DI actually. And considering you can see the quality difference between 35mm and Super-16 even at standard def video resolution, like on broadcast TV (hence why most high-end commercials are shot in 35mm), you can definitely see the difference at 2K.

Edited November 5, 2004 by David Mullen

[Daniel Henriquez Ilic]

Most theatrical films still don't use a DI actually.  And considering you can see the quality difference between 35mm and Super-16 even at standard def video resolution, like on broadcast TV (hence why most high-end commercials are shot in 35mm), you can definitely see the difference at 2K.

<{POST_SNAPBACK}>

 

I agree, but now thanks to 2k DI blow up from s16mm to 35mm, one is able to get sharper 35mm finish, in relation to conventional optical blow up.

[Guest Daniel J. Ashley-Smith]

Most theatrical films still don't use a DI actually

How do they edit it then? The old splicer and cement?

[David Mullen ASC]

They usually transfer the negative (camera rolls) daily onto tape (on a cheaper telecine to an NTSC or PAL tape format), digitize the footage into a computer, edit it digitally, generate an EDL (Edit Decision List) with time code and, more importantly, film edgecode (keycode) information, then hand the camera rolls to a negative cutter, who physically cuts and cement splices the film, matching the EDL. Then the cut negative is answer-printed, etc.

 

A color-timed interpositive is made off of the negative for a final high-quality transfer to video (usually HD these days) for making various home video, cable TV, and broadcast masters.

 

For a DI, it's usually done in the same manner except that after the EDL is created, they go back and scan the select shots off of the camera rolls into a digital file format, create an edited digital master (sort of like an online, assembling the cut version following the EDL numbers), digitally color-correct, do a dust removal session, and then record the digital data onto a 35mm negative or internegative. After that, again, the same steps: answer printing (although that's pretty simple because the movie is color-corrected already), etc. The digital master (usually 2K) is used to make HD masters, and then PAL and NTSC masters, etc.

[Eric Steelberg ASC]

There's a good article about the process here .

 

An important aspect to getting a good DI, other than people who know what they are doing, is using an actual scanner and keeping the scan as data to retain the full bandwidth of color...not putting it out to tape.

 

I'm just starting to pep a <$4mil film and we're doing a lot of research on the process. The above article also talks about more cost effective ways to do an HD DI...whick is really just an HD finish put back to film.

[Wendell_Greene]

The French lab Eclair also seems to produce good results.

 

 

Yes. I'm looking forward to seeing the results of Rodrigo Prieto's work with Yvan Lucas of Eclair on the DI for "Alexander".

Edited November 5, 2004 by Wendell_Greene

[Max Jacoby]

Yes.  I'm looking forward to seeing the results of Rodrigo Prieto's work with Yvan Lucas of Eclair on the DI for "Alexander".

I did not know they used Éclair on 'Alexander'.

 

Yvan Lucas is a legend, he does all of Kondhji's work among others.

[Max Jacoby]

I agree, but now thanks to 2k DI blow up from s16mm to 35mm, one is able to get sharper 35mm finish, in relation to conventional optical blow up.

I beg to differ. I have seen some very good Super16 Optical blow-ups recently and I believe you can get a better result (i.e. sharper, with better contrast and colors) than doing it digitally.

[Filip Plesha]

Here is my opinion:

 

The problem is not low bit deph. In still photography 24-bit color (8-bit per chanel) is used as a standard for printing. Usually scans (and digital camera files) are captured at more (16-bit per color is the gold standard), but after every correction has been done, it is converted into 24-bit for printing. The results differ, but best laser machines loaded with state of the art scans create prints that match optical prints (at least) or make even better prints than good opticals. And all that at just 8-bit per chanel. No fake skin tones, no plastic colors or whatever. Top of the line images.

 

I think the problem is in the recording technology and in resolution.

Digital technology is not a problem on its own. For every shade and nuance on the negative, however subtle and delicate it may be, there is a numerical value.

And I think scanners (specially those high end MP scanners) do a very good job of

collecting information from film.

I belive it is a llittle bit of everything. first of all it's the 2K resolution, then it's grain removal done at this low resolution, then it's too much grain removal in general, then it's perhapse unprecise film recorders.

These problems will most certainly dissapear in time, no need for pesimism.

[Daniel Henriquez Ilic]

I beg to differ. I have seen some very good Super16 Optical blow-ups recently and I believe you can get a better result (i.e. sharper, with better contrast and colors) than doing it digitally.

<{POST_SNAPBACK}>

 

I have seen excellent s16mm optical blow up to 35mm (through 35mm IP and IN) done recently at Cinecolor in Argentina. I understand what you mean : staying in pure photochemical world without sampling and bit depth issues... but I have seen impressive s16mm blow-up to 35mm through 2k DI, with beautiful color, good contrast and sharp.

[Daniel Henriquez Ilic]

Here is my opinion:

 

The problem is not low bit deph. In still photography 24-bit color (8-bit per chanel) is used as a standard for printing. Usually scans (and digital camera files) are captured at more (16-bit per color is the gold standard), but after every correction has been done, it is converted into 24-bit for printing. The results differ, but best laser machines loaded with state of the art scans create prints that match optical prints (at least) or make even better prints than good opticals. And all that at just 8-bit per chanel. No fake skin tones, no plastic colors or whatever. Top of the line images.

 

<{POST_SNAPBACK}>

 

Yes it is true (but hard to beleive) that high end digital printer Durst Lambda have only 8 bit per channel color bit depth output.

[Phil Rhodes]

Hi,

 

8-bit is fine so long as you don't want to change it at all. If you want to grade it, well, 10-bit log is noticeably better.

 

Phil

[Filip Plesha]

Hi,

 

8-bit is fine so long as you don't want to change it at all. If you want to grade it, well, 10-bit log is noticeably better.

 

Phil

<{POST_SNAPBACK}>

 

yea, that's exactly why photo editing is done at 16-bit. After it's done

it's back to 8-bit.

[Guest Sean McVeigh]

Besides the lack of resolution compared to contact printed 35mm

<{POST_SNAPBACK}>

 

I haven't purposely viewed the two side by side, but I remember reading recently that Fellowship of the Rings was a mix of 2K DI (for all the effects shots) and good old-fashioned contact printing. Apparently Peter Jackson wasn't entirely happy with the way the two looked when intercut so the remainder of the 3 films were finished entirely at 2K.

 

Those films looked plenty good to me in the theatre.

 

But yah, nowadays most films that are going DI are going 4K. I recently spoke to someone at Technicolor about this. Even 4K downsampled to 2K is preferrable to 2K.

 

I've also read some accounts arguing in favor of DI citing that after the myriad of contact prints required to get to a final projection print (IP/IN/...) often you are left with something akin to 1K. I find that a bit hard to believe, but it's certainly not outside the realm of possibility.. of course the nice thing would be that the generational resolution loss through analog printing is probably preferrable to nice blocky pixel artifacts.

[Michel Hafner]

For the grading itself there seems to be a trend in American DIs to use grain reduction a lot, which will make the picture very smeary looking. 'Kill Bill' and 'Big Fish' are some of the worst examples in that regard. Somehow the Europen DIs I have seen (mainly French and English) tend not to do this as much. There grain is not seen as something one should try to get rid off at all costs.

The French lab Eclair also seems to produce good results. On the other end of the spectrum there is EFilm who do a lot of films, but who have failed to impress me so far. They seem to be in love with DNR and the results  can be quite appalling at times.

<{POST_SNAPBACK}>

 

Good to see that someone else hates the DNR look of so many DIs and speaks out against it.

DNR totally undermines the beauty of the original photography. It's a plague.

The worst I have seen so far was "Seabiscuit". But it's in so many films now, Lord of the Rings trilogy too.

Depressing. :(

[David Mullen ASC]

Actually E-Film does very LITTLE grain reduction and have a philosophical bias against doing it, which is why you sometimes see some shots are grainier than other shots in their D.I.'s because it's that way it was in the original photography. In fact, for a long time, their color-corrector did not even offer DNR. Someone told me this after I asked why that underexposed scene in "XXX" during the bridge stunt wasn't degrained a little to match better.

 

Same with film-outs from HD masters -- E-Film does less noise reduction processing than many other facilities unless you ask for it.

[Michel Hafner]

Actually E-Film does very LITTLE grain reduction and have a philosophical bias against doing it, which is why you sometimes see some shots are grainier than other shots in their D.I.'s because it's that way it was in the original photography.  In fact, for a long time, their color-corrector did not even offer DNR.  Someone told me this after I asked why that underexposed scene in "XXX" during the bridge stunt wasn't degrained a little to match better.

Same with film-outs from HD masters -- E-Film does less noise reduction processing than many other facilities unless you ask for it.

<{POST_SNAPBACK}>

 

People must ask for it a lot then. Often when I see the DNR look and check the end credits it says DI by

'you know who'. But other facilities do it too, no doubt. Eclair did it on "Two Brothers".

[Michel Hafner]

I think the problem is in the recording technology and in resolution.

Digital technology is not a problem on its own. For every shade and nuance on the negative, however subtle and delicate it may be, there is a numerical value.

And I think scanners (specially those high end MP scanners) do a very good job of

collecting information from film.

I belive it is a llittle bit of everything. first of all it's the 2K resolution

<{POST_SNAPBACK}>

 

2K has got nothing to do with plastic skin. It refers to spatial resolution only. I have tons of DVDs with << 2K which have no plastic skin on display.

Fake looking skin concerning color must be the result of bit depth per sample, rounding issues during grading and color space mismanagement during display and film out calibration.

Fake looking skin concerning texture and motion is caused by DNR and 'grain management' on the other hand. Skin offers not many typical features for the motion estimation to hold onto and motion vectors are usually off and create an ugly chaos during temporal filtering.

[Filip Plesha]

2K has got nothing to do with plastic skin. It refers to spatial resolution only. I have tons of DVDs with << 2K which have no plastic skin on display.

Fake looking skin concerning color must be the result of bit depth per sample, rounding issues during grading and color space mismanagement during display and film out calibration.

Fake looking skin concerning texture and motion is caused by DNR and 'grain management' on the other hand. Skin offers not many typical features for the motion estimation to hold onto and motion vectors are usually off and create an ugly chaos during temporal filtering.

<{POST_SNAPBACK}>

 

 

I didn't say that low resolution alters colors, but softer digital image projected in cinema has a unique apperance because we are all used to grain.

Before DI or HD, if you had a soft image it was grainier that usuall because you were either seeing 16mm or a bad grainy duplicate of 35mm negative.

But nowdays you get the softness, but without additional grain, so it looks

unusual, which is (I think) a part of that DI look that everbody keeps talking about.

 

But again, I can assure you that bit deph is not the problem here on its own.

The whole photo industry from breathtaking high end prints to consumer photos is based on 8-bit graphics. The only thing that makes some of these prints better or worse in terms of color is the equipment (scanners, printers), not the 8 bit standard. Cinematography uses 10 bit which is much better, and on top of that it's logarythic, so its a better usage of 10-bit space.

On it's own it is quite sufficient, but It does have a limited headroom for manipulation. For really wild manipulations perhapse 12-bit or even 14-should

be a better choice. But output and storage can still be 10-bit

 

And one more thing. I believe, it's not really so much color deph that makes that organic film look. When you look at some 50's eastmancolor duplicate negative , you are not really seeing millions of soft gradations in the image (like you would se on today's Provia for example).

Instead what you see are rather simplified rough colors with not many fine gradations, but you still see this film as organic, textural and analog looking.

In fact I've seen some examples where a DSLR would make finer gradations

and more color nuances on some areas of the photo, and some older film would avarage out and flatten these colors, make them more cartoonish, but still film version would look more film-like and more organic. It's hard to explain.

Which brings us back to the eternal question of what makes film look like film.

It's not resolution, it's not color deph, it's not even latitude (blown out skip bleach film still looks organic and film-like) it seems it is not any of those things, it's just

a different way of reacting to light.

Edited January 7, 2005 by Filip Plesha

[K Borowski]

The whole photo industry from breathtaking high end prints to consumer photos is based on 8-bit graphics. The only thing that makes some of these prints better or worse in terms of color is the equipment (scanners, printers), not the 8 bit standard.

<{POST_SNAPBACK}>

 

I'm not sure how you do it, but my photographs, be they still, or shot at 24 fps, are on film. There are no "bits" just grains, and I am not by any means alone in that I shoot exclusively on film. There are people in still photography who shoot on 8x10 inch film and pay roughly $20 PER PICTURE. I know that many fine art photographers would rather have their pictures burned than reduced to 8 bits. Magazine photographers regularly shoot on 6x6 or 6x7 cm roll film. There is a hell of a lot more information in a still negative than the best DSLR (unless you have a camera that needs to be tethered into a wireless network and takes ten seconds to take a "picture"), unless you're were into shooting that 8- or 1600 speed crap before you joined the "digital revolution" that's sweeping the world and making Asia a lot richer. Please don't assert that the "whole photo industry" is based on 8 bit graphics. A lot is still based on Eastman Kodak photochemical innovation. It irks me to no end when people assume they speak for an entire industry, especially one that I anticipate being involved in (shooting film all the way to the end) for a long time without the need for digital manipulation or technology.