Cinelog - True logspace conversion for DNG and CinemaDNG footage

[Andy600]

@Danne - we don't include a batch script. The best option is http://aescripts.com/batchprocessor

@Mageye - we are working on a basic workflow video but if you use VisionLog you will already be familiar with how things work. How you grade the image, and the NLE you use to edit, is entirely up to you but we provide LUTs to get your image into Rec.709 colorspace plus several 'looks'.

A basic After Effects workflow would be to convert your raw files to DNG, import your sequences to AE and apply Cinelog in ACR. You really only need to tweak white balance (if needed) and click OK. Then add the sequence to your timeline, set the correct frame rate and export to a high-end codec (Prores HQ, DNxHD, Cineform, Grass Valley HQX or DPX etc). You can then work with video files that have all the latitude of the raw files.

The difference between Cinelog and Visionlog might not appear big at first glance but in a professional workflow, where retaining quality throughout post production is vital, it can be significant, especially where highlight and shadow information is concerned. VisionLog (from what I can tell) is not logarithmic and is not a mathematically calculated curve, which means the steps between f-stops are not uniform and smooth. Shadows and highlights are squashed at each end with the mids almost linear (it's a bit odd). It also lifts black levels a lot (a good trick to hide noise) and if you render intermediate videos like that you will inevitably lose some latitude and detail when trying to open up the upper and lower ranges of your image. If you want to work with raw footage, as raw footage, you are better off working linear (i.e. no curves applied) but that can be tricky. Obviously realtime playback of your raw footage is very CPU intensive and workflow is very very slow because of debayering + applying plugins etc so that option isn't usually viable for anything over a few minutes long.

The other significant difference between Visionlog and Cinelog is that we embed each camera's individual matrices and most other metadata, specific to the model. Cinelog is a dual illuminant profile and ACR uses this to interpolate colors depending on the white balance settings to help keep colors accurate (it helps with difficult lighting situations). We also tweaked the standard DNG colors in several places to fix some issues with Canon colors so that skin tones are where they should be (have you ever noticed what DNG conversion does to skin colors on Canon images?). Anyway, I could go on and on but you want to see results so keep an eye on the website for some interesting things

[iaremrsir]

Already tried with linear profile DNGs and Cinelog embedded in a LUT. The results are not as good or as accurate as you get with the ACR workflow. You can get fairly close with a LUT if the embedded DNG metadata is correct but Resolve only uses the first DNG color matrix and it's DNG implementation is quite basic. We had issues with color casts, skin tones and other things that just don't happen with ACR plus ACR still beats Resolve's debayering 99% of the time.

We have a couple of pros putting some examples together and the differences are quite noticeable. If you use a batch script for rendering DNGs with Cinelog through AE it's comparable in speed to rendering intermediates from raw files in Resolve. I will post everything on our website when the content is ready.

Well that's disappointing to hear  But it's good to hear you guys already went the extra mile and tested with another program. I really wish Adobe would find a way to offload the debayering and processing. And not make us confirm settings for every DNG clip we import.

[kgv5]

Cinelog is a mathematically calculated log curve i.e. it's very accurate (it's similar to Log-C from the Alexa). It doesn't crush shadows, does not add any contrast to the image and retains all of the DR of your raw image. Cinelog also has full implementation of all color matrices useable by the cameras. The additional Cinelog CT profile also has a built-in LUT that improves on the standard DNG color palette and gives great skin tones on Canon cameras. Plus there is a nice LUT pack included

The Visionlog profiles crush shadows and lift the black point to make the image appear flat plus the gradations between f-stops are not equal. It uses a custom curve and isn't logarithmic. You might think this doesn't matter as it's still raw data but it matters a lot if you want to make the highest quality intermediates for color grading or inter-cut with log footage from different cameras.

So, does this mean that we are recovering all the details from the highlights and shadow just like putting ACR highlights and shadow sliders to the +/- 100 value?

[Andy600]

@iaremrsir - Yes, that would be nice. Hopefully Adobe will make ACR more video friendly but I can't see that happening anytime soon.

@kgv5 - Yes for detail but the ACR highlight and shadow sliders control intelligent highlight and shadow 'reconstruction' which is basically a form of local contrast, and very similar to tone mapping. Unfortunately, ACR is too clever sometimes. It applies some settings dynamically, depending on the specific image it is processing, and if the next frame in the sequence has a significant change in luminance ACR will adjust dynamically and this can manifest as flickering when you play back the video. Cinelog basically negates the need to touch those sliders. You do, of course, need to render in a good codec to get maximum latitude for grading but you can pull back as much information as ACR gives you (perhaps even a little more if you use tone mapping). Cinelog itself will not give you a 'HDR/Tone mapped' look but does keep all the information intact.

[cpc]

Just a quick correction: Resolve 10.1 most certainly uses both the color matrices and both the forward matrices. I haven't bothered checking with previous versions.

[Andy600]

Just a quick correction: Resolve 10.1 most certainly uses both the color matrices and both the forward matrices. I haven't bothered checking with previous versions.

Thanks for clarifying

Just to add to that: The matrices need to be embedded in the DNG for Resolve to utilize them and (afaik) most ML raw converter apps currently either use ML's default single ColorMatrix1 (unique to each model) or just duplicate it (i.e. Raw Magic) for the second. Maybe it's different if working in ACES with input transformations? but Resolve only has them for the 5D (doesn't say if its Mark 2 or 3), 7D and 1D.

DCPs can override this so even if the matrix/matricies are set incorrectly the profile's settings will be used (if selected, and if the DNG has the correct UniqueModelName tag). I love working in Resolve but footage (especially ML raw footage) just looks better when it's been debayered in ACR.   

If Blackmagic added DCP support we'd be halfway there.

[kgv5]

Andy600, thank you for the answers, i have bought it and will test in the next days.

[cpc]

Maybe it's different if working in ACES with input transformations? but Resolve only has them for the 5D (doesn't say if its Mark 2 or 3), 7D and 1D.

Canon cameras ACES IDTs in Resolve are for camera native codec footage (although unclear to me what picture profile is assumed).
As the DNG tags should technically describe the camera space fully, any DNG footage is imported with the CinemaDNG ACES IDT, no matter what camera it comes from. ACES workflow in Resolve is still a bit rough though.

[mageye]

@Andy600

Thank you for your reply it's informative stuff. You say that once converted 'you will have the full latitude of the raw files'. If this is true it's excellent. Excuse me for being cautious, but seeing as things move quickly and possibly there will be another way of treating the images in a few months, I would personally like to keep backups of my RAW files (just in case!). Again, I would really like to know rather than think I know that I am retaining the best quality image. It always seems that almost any conversion will have it's caveats. I apologise, it's just my suspicious nature!

However, I am really interested in what your profiles have to offer and I will indeed be scrutinising your website for any information that may convince me. Of course I can't wait to see some real world comparisons and examples. I am nearly there!

Thanks again.

[Andy600]

@mageye - Nothing wrong with caution. I'm the same myself

If you want to have a play with an image or two just send me a couple of your DNGs and I'll convert to them Cinelog as TIFFs. Incidentally, we're opening a private LUT bank for Cinelog users later this week and we'll be adding new content weekly. We have a couple of interesting conversion LUTs to add (Cinelog to Log-C colorspace and Panavision Gensis Log colorspace) plus a couple of new 'look' luts including one that replicates the color gamut and gamma in Act 3 of Skyfall.

[N/A]

Very interesting, can't wait to try them out! Been doing more work lately with raw for music videos, and now that my workflow is pretty much down pat, I want to start testing more grading options.

[bnvm]

@Andy600 If I am going out to a high bit depth intermediate like 16bit, is there any benefit to using your log profile?

[Andy600]

@Andy600 If I am going out to a high bit depth intermediate like 16bit, is there any benefit to using your log profile?

That depends on how much flexibility you want in post, how much hard disk space you have or need, how big the project is, are you handing over to a colorist etc. What codec/image format are we talking about exactly?

Log will give you the benefit of a more film-like response in the shadows/highlights. Given that using ACR's recovery sliders can cause flickering if used your only option would be to stay in the raw domain until final output as you would need to grade or at least balance each clip as you go, that's a very slow way to work. If your projects are not large or time critical then working with raw files is an option (if your DNGs are meta-tagged correctly so ACR can understand what the colors are really supposed to look like).

With Cinelog you can literally import your footage, check the WB (or just use 'as shot' WB if you're confident of your shoot settings), set the frame rate and render to an intermediate without having to think about shadows and highlights until later i.e. you can trust it. It's the equivalent of having Log-C Prores HQ or 444 files to work with but you can decided on the codec depending on your needs and/or how important the footage is. We do still recommend output to Prores HQ or higher but that's up to each user. I personally output to 10Bit DNxHD most of the time and I'm happily now deleting my raw files and freeing up TB's of space. I'm not sat thinking "could I have got more from the DNGs?" because Cinelog retains the highlight and shadow information, the colors are great and everything is a whole lot faster.

[bnvm]

I guess what I am really wanting to know is would exporting a 16bit tiff file using your log profile offer any benefit over just exporting a 16bit tif without it. I realize if I decide to go with a 10bit intermediate then it is obviously worth it but for 16bit it may not be unless the profile working during the debayering process increases the useable dynamic range for example.

I am going to be doing some vfx work and having a log encoded image presents some issues that require some extra steps that may not be worth going through if I decide to go with 16bit intermediates. So far I have been getting pretty decent performance using 16bit. Now if the log encoding increases dynamic range then going with a 10bit log and dealing with those issues might be worth it.

[iaremrsir]

Unless you're going to uncompressed linear, 16 and 32 bit files are unnecessary. Logarithmic curves were made for compression, which means they're for 8-12 bit files. Log encoding doesn't increase dynamic range, it remaps how it is stored.

[Andy600]

@bnvm - Yes, as iaremrsir says, it will not increase your dynamic range. The DR is governed by the sensor.

There is a misconception about 'log' and the term is often abused. To most film makers, log is simply a flatter image from raising the shadows and lowering the highlights but it's actually very important to image quality and color grading to properly calculate it. This also includes understanding of how log curves work in relation to the dynamic range of the raw image and the application applying the curve. With the BMCC we had to calculate a slightly different curve, taking into account it's 13 stop dynamic range and the ACR workflow. Log will not truncate the data. It efficiently compresses a linear source into a 'container' with an equal number of steps between each f-stop so the lower stops are not stretched in relation to the high stops (i.e. better shadow detail) and so long as you render this to a good codec you will have maximum latitude for color grading. It's because of this that a true log image can be pushed further in post than a 'pseudo-log' flattened image.

Re: vxf - I think you can set the working space individually for any clip in AE in the interpret footage panel? If you do this and linearize the workspace you shouldn't have big problems working with log footage. We don't have an ICC profile for Cinelog (yet) but the Universal Camera Film Printing Density color space should work ok. There is a tut somewhere online about composting with Alexa Log-C footage and the same thing will apply for Cinelog.

[bnvm]

Thanks that is the answer I was looking for and what I suspected. I just wasn't sure if there was something in the way ACR works that in some way limited the dynamic range of the output, sometimes there are compromises that are made for reliable output/color reproduction, were the log curve might retain a little more.

Thanks for answering my question.

[Andy600]

@bnvm - here's that vfx/log tut I was referring to http://wicuslab.com/working-with-arri-alexa-logc-footage-in-adobe-after-effects/

[bnvm]

Thanks for the link I will check it out.

[dossisman]

Recently purchased it. I'm still using VisionLOG. I felt that while Cinelog obtained more details in my mids, VisionLog preserved my shadows a LOT better.

With that said... I am TOTALLY in love with their Film Gamma 2 LUT. Especially when I apply it to my footage with VisionLOG.

[Andy600]

Recently purchased it. I'm still using VisionLOG. I felt that while Cinelog obtained more details in my mids, VisionLog preserved my shadows a LOT better.

With that said... I am TOTALLY in love with their Film Gamma 2 LUT. Especially when I apply it to my footage with VisionLOG.

Really? In our tests VisionLog tends to crush shadows significantly and gamma is off. This is workable if you're still in the raw domain but once you go to an intermediate format you'll struggle to get anything useful back without introducing a lot of noise. Also, we measured black, white and grey points and they don't hit the mark without a lot of push-pull in post. It won't be hugely noticeable if you're just outputting for the web but for film or Rec.709 HD TV conformity you need accuracy (especially for the BBC).

You might still get good results with VisionLog if you stick to using their premium LUTs (BTW, they do make some nice looking LUTs from what I've seen) and this is understandable as they are developing their own system but the professional colorists we converse with all prefer to grade without LUTs (except when pushed for time) and for this they much prefer an unaltered linear or logarithmic base. They also work to strict standards for broadcast work and just because something may look nice on screen doesn't mean it meets these standards. There are limits to luminance and saturation values amongst other things that need to be adhered to. We aim to help colorists get to final output quicker and with less work. 

BTW You should check out the forthcoming free update to Cinelog. I think you'll be suitably impressed

[morsafr]

A quick question about Cinelog usage: is the profile optimized only for the "2012 process" or is it also compatible with "2010 process"?

I'm wondering because in some cases the 2010 can help avoid flickering.

Thanks for the feedback!

[dubzeebass]

A quick question about Cinelog usage: is the profile optimized only for the "2012 process" or is it also compatible with "2010 process"?

I'm wondering because in some cases the 2010 can help avoid flickering.

Thanks for the feedback!

The point is that touching the sliders is moot with Cinelog as I understand it. You just white balance and then export.

[morsafr]

Indeed that's the point of Cinelog.

But as far as I understand the 2012 process is adaptative for each image contrary to the 2010 version.

[Andy600]

The ACR process method will not affect how Cinelog works but there are subtle variations to gamma and detail between 2003, 2010 and 2012 processes. It's entirely up to you which one to use. I prefer 2012 personally.

WB is the only adjustment I would say is required, even just to fine-tune it but making changes to exposure, using CA correction and Defringing highlights is ok. You can use sharpening too but I personally prefer to use an unsharp mask (0.6 radius / 50-120 amount) or a HPF. ACR Noise Reduction is ok to use but slows render times down. Neat video used on intermediates is a better option if you need NR and is much more controllable.

We're just getting ready to release a big update to Cinelog including a new preview profile so you can check exposure and WB without leaving the ACR interface. There are new, accurate Rec.709 LUTs and one for VFX artists who work in Linear 1.0 gamma plus some 'approximate' colorspace conversion LUTs to put Cinelog footage into similar colorspaces of some well known high-end cameras plus a few other goodies. It's a free update for Cinelog users.

I'm just finishing the updated manual before release (hopefully ready today/tomorrow).