Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - hyalinejim

General Chat / Re: 5D III used on The Mandalorian
October 09, 2020, 01:03:00 PM
Haven't watched the vid, but presumably it was stills?
Did you use diffusion on the Ifbagon or is that a result of the lens itself? It's very nice.
Thanks! Keep us posted on what you uncover :)
General Chat / Re: Panasonic GH5 or Canon 5D Mark III?
October 11, 2018, 10:05:04 AM
I switched to GH5 after using 5D3 ML for a few years.

If you're doing jobs for clients the GH5 is a no-brainer in my opinion because of smaller file size, 4K and IBIS. If you are more casual or are doing a passion project then 5D3 ML is absolutely marvellous.

VLog colours are crappy out of the box but the 10bit codec will stand up to extensive grading and you can get lovely colours from the GH5. Here's a comparison I did in the early days of the GH5 where I was trying to match its look to the look I was using on the 5D (since then I use a different look, but it shows that the GH5 isn't limited to Panasonic's default colour):
Very interesting!
Hi Ibrahim, and congratulations on your first short!

In terms of grading, generally it looks good. However, I think that your blacks are lifted too much. I know that this look has been popular over the last ten years or so, but I think it makes your film look too washed out.

However, this might be a problem with your export settings, as the RGB values for the letterboxed area (which should be 0, 0, 0) is actually 16, 16, 16.

I think I see some sharpening with a wide radius or perhaps it's a local contrast adjustment... I would be inclined to not do that. It's particularly crunchy looking at around 4:15

There is also some colour banding on skin (for example at around 5:55). Maybe you are using a LUT that is causing this.

Finally, I think there is a frame rate problem somewhere in your workflow as some clips are a bit jerky.

Aside from these grading considerations, the camerawork is good and so is the editing.
I used Cinelog for this one, chief! Thanks again to Andy for an awesome DCP profile.

It's been a while since I shot ML RAW as I'm using a GH5 a lot these days, but if I was doing it from scratch I think I would use ML Log. IIRC, the great promise of ML-Log was that colours could be very accurate with one of the versions (but there were odd artifacts). The version without the artifacts had acceptable, but not 100% colour accuracy.
Maybe try it in the other folder and see if it works.
My short documentary The Cloud of Unknowing, shot on 5D3 Magic Lantern, has been nominated for "Best British / Irish Short Film" by the London Film Critics' Circle. The awards ceremony is on January 28th... so fingers crossed for that!

Here is a short extract from the film with some info about it:

And here is more details on the award nomination.

Thanks Magic Lantern!  :)
Quote from: 70MM13 on November 26, 2017, 04:16:41 PM
This looks fantastic!
Please port it to resolve!

Try this set of LUTs for a similar look in Resolve:

I used these settings, so it's expecting this kind of input:

Then you'll need to use curves or lift, gamma, gain to bring the log signal into range:

Quote from: beauchampy on November 28, 2017, 05:28:34 PM
Any chance you could do this for the c200?

In theory, yes. I would just need a C200 first! But there's a fair bit of work involved in making these luts. That's something that I might start doing commercially for cameras other than ML over the coming months.

Yes, I considered different scan options but decided to stick with a lab scanner as they impart a look of their own. The chart shots were processed and scanned on a Noritsu by Carmencita Film Lab. They have a good reputation and some of the work on their website is amazing.

If these luts do what they're supposed to, it should be possible to get close(r) to these kinds of looks:

Yes, I'll do that at some stage when I feel that I've got it nailed. There are a few variables that I still need to figure out, and I need people's feedback here.

I'm going to upload an alternative set of the five luts. In this second set, the exposure does change with each lut. So the underexposed luts will give you a darker image than the overexposed ones. I'm curious as to which of the two sets people prefer, and why. I think I might prefer the second set as I like seeing the difference in exposure:

Another question is how to manage the colour casts. In both sets, I've been calibrating the white balance to 90% white for each exposure level. This more or less standardises the colour across all five luts. Then I go back and make "C" (the properly exposed shot) neutral across most of the midtones. When I apply this correction to the other four it re-introduces some of the interesting colour casts, which I think is desirable. The problem is that each film scan (on a Noritsu or Frontier film scanner) can vary a little from shot to shot. It's not like opening a series of shots in ACR where, if they were shot with the same settings, there will be absolute consistency. So in designing these luts I need to figure out when, where and how to compensate for the variances inherent to film scans. The luts made directly from the scans are a little bit inconsistent in the progression of colour and contrast changes that you would expect.

In pictures, here's the difference between the two sets:

v1.0 (exposure stays more or less constant)

v1.1 (exposure actually changes)

One final thing to point out (and this is something I may need to work on) the underexposed versions of the luts give odd shadow colour and tonality if you pull the blacks down before the lut. The reason for this is that the RGB curves diverge very strongly in the shadows for the underexposed luts. If you want darker shadows with these versions it's better to pull them down after the lut.

On the other hand, all five luts loooove overexposure. So don't be afraid to lift the midtones pre-lut for a very contemporary effect. There should be good highlight roll off too!
...[fanfare of trumpets]

I'm ready for you guys and girls to check out this set of Ektar 100 Luts:


These are built for ML-Log v1.3, which is a work in progress.


Read more about ML-Log:

So you can think of these luts as also being a work in progress. They are designed to be applied in the following workflow:

1. Import DNGs into Adobe After Effects
2. When Adobe Camera Raw opens, choose ML-Log v1.3 for the Camera Profile
3. In After Effects add the Apply Color Lut effect, or export to an intermediate codec for grading elsewhere

What's the deal with these luts?
These luts fairly accurately emulate Ektar 100 film, as scanned by a specific lab and at various exposures. With negative film as exposure increases so does colour and contrast. And there are different colour casts in the shadows, midtones and highlights at different exposure levels. The five luts here emulate these changes in colour and contrast from two stops underexposed to two stops overexposed. I've labelled them A to E and added plus or minus marks to indicate the exposure level. I've also carefully adjusted the tonality and white balance of each lut so that middle grey falls at the same point for each one. Applying the minus two lut won't make your footage look underexposed, nor will the plus two make it look overexposed. However, you'll notice that the minus two has very milky blacks, a generally cool tone and muted colours. The plus two is much more contrasty with overall warmer tones and hyper saturation, especially in the reds. It may be too much for skintones, as is often the case with real Ektar 100 film. The middle exposure lut has the most neutral colour cast.

How to approach the luts in post

There are two things you're going to want to be able to do to work with these:
1. Manipulate white balance somehow (both on the orange-blue axis and red-magenta axis). Lumetri is good for this. You can also try a three way colour corrector in the midtones
2. Manipulate black, midtone and white levels. I like to use curves, but levels or lift-gamma-gain controls are also good

Both of these should be done before the lut. So don't feel that you're stuck with a particular colour cast or contrast level with a given lut. If you like the colour of "D +" (one stop over) but find it too contrasty, just lift the black point and lower the white point before the lut to lower overall contrast. Use the midtones control to adjust overall exposure.

I 'll illustrate the differences between the five of them in more detail later. For now, here are some random CR2s I grabbed from Imaging Resource and applied the luts to. These images were exposed well and the results here are straight out of the box:

MLLog1.3 to Ekt100 A - -

MLLog1.3 to Ekt100 B -

MLLog1.3 to Ekt100 C

MLLog1.3 to Ekt100 D +

MLLog1.3 to Ekt100 D + +

If there's interest, I'm pretty sure I can do up versions for Resolve as well.
I confess I know as little about GPL licensing as I do about code! As Danne mentioned, there was no coding involved in making this profile just the use of some free and paid tools as outlined in the opening post. I'll remove the reference to GPL in the first post for now.

Has anybody been stress testing the profile? Danne and I have noticed some colour artifacting under certain conditions. Look at the colour fringing around patch 4E:

Digital Colorchecker SG CR2 downloaded from Imaging Resource. Temp and tint sliders at the extremes in ACR. ML-Log v1.0

This can be prevented by choosing a different Tone Reproduction Operator (TRO) in Lumariver (ML-Log v1.0 uses "ACR" TRO and v1.3 uses "Simple")

ML-Log v1.3

All well and good, but the colour accuracy of this profile is slightly affected.

ML-Log v1.0 - the inner square is the correct colour the outer square shows the result of the profile

ML-Log v1.3

With v1.2 reds, yellows and purples are brighter but it might hurt colour accuracy. Can you guys try to stress test this new version versus v1.0? If v1.2 still has good colour for real world images it's the obvious choice as there are little to no colour artifacts with it.


Very exciting times ahead for me and for fans of this lut: today I got back scans of the first roll of Ektar that I've shot. I took a series of chart shots at different exposures with which to do the matching, and I also had shot some A/B comparisons with Ektar 100 loaded into my OM1 (nice little camera!) and regular old CR2s on our friend the 5D. I spent a few hours this evening working on matching them. Some shots match better than others, that's for sure, as there are a lot of variables at play when shooting film. But this is one situation where the lut makes the two cameras quite close. The base for this lut was the new DCP profile that Danne and I worked on: ML-Log

OK, here's the 5D shot with ML-Log profile applied in Adobe Camera RAW:

Now, here's the Ektar 100 shot. Nice, isn't it?

And finally... here's ML-Log with the Ektar matching lut applied, with some exposure and white balance tweaks. Also nice!

And one more for good luck! The film scan is in the middle:

I'm gonna work on this some more over the next week or so and when I'm fully happy with the results I'll release it.

Try Juan Melara:

Or anything designed for Cineon or RedLogFilm will hopefully work well.
@bpv5P I hope you find it useful.

Everybody who tries this, please let us know if it works out well for you. We're going to use this v1.0 5D3 profile to fine tune anything that might need fine tuning before attempting to create similar profiles for the other cams.
Ideally, it's best to do it first, Lars. But you can also change it afterwards if you have a good white balance tool. The Lumetri color panel is available in AE and will work.

But I wouldn't agonise over getting it exactly perfect - you'll be able to push the white balance quite hard in post. Sometimes you're not sure what exactly the final WB should be until you're ready to grade and that's fine.
In brief, this is the workflow:

ACR defaults:

Apply ML-Log DCP profile:

In After Effects (or Premiere, or Resolve) add curve or lut and tweak exposure

(5D3 only for now)

It gives me great pleasure to announce the imminent arrival of ML-Log, a DCP log profile for Magic Lantern DNGs. Those of you who enjoy the benefits of using Adobe Camera Raw over other methods (superior debayering, colour rendition and noise control) might like to give this profile a whirl.


Key features:
1. Maps RAW linear to a logarithmic curve with blacks at 9.3 IRE, 18% grey at 46 IRE and 90% white at 72.6. This gamma is very close to Cineon or RedLogFilm

2. The profile is fairly colorimetrically accurate. All you need to do is apply a curve to restore proper contrast and colours should look good. In this screengrab the outer area of the patches is a CR2 shot of a ColorChecker with ML-Log applied, followed by a curve to match the gamma to that of the chart values. The inner squares are an RGB visualisation of a typical ColorChecker. You can see it's not a perfect match, but it's close.

3.  I've included two options for de-logging footage based on an old paper on the relationship between Cineon and film. One of these curves has a colour cast, and the other is clean. They are included as Adobe curves, and as .cube luts. That paper is here, and it is interesting You should get some nice results with this, but do try other luts as well.

How the profile was constructed:
With Danne's help, I used LutCalc to generate a linear to log curve. I then incorporated this into Lumariver Profile Designer, which is a GUI for DCamProf - a DCP designer. The matrix and HSV tables were derived from a CR2 shot of an X-Rite Colorchecker using Lumariver's inbuilt colour calibration function. If you are interested in seeing what's going on inside you can use DCPTool to peek at the contents.

What's next?
I'd like ML users to test this profile for usability. Any comments, questions and criticisms are very welcome. If it proves useful, we should be able to generate similar profiles for other Canon cams. Not only that, but there's no particular reason why we need to be locked into this particular curve. It should be possible to generate curves to match, for example, Sony's S-Log, Canon's C-Log, Panasonic's V-Log or Arri's Log-C

So will this turn my camera into Alexa / C300 / Varicam / etc.?
That would be nice but, no. This profile does two things: it changes the gamma of the RAW output so that it matches a particular curve (in this case Cineon / RedLogFilm) and it slightly tweaks the colours so that they are more accurate. The advantage of this is that you can apply this profile to your DNGs, render to a lossless intermediate codec, save on hard drive space, make editing smoother and defer decisions on colour correction until after editing. But it won't make your Canon look like an Arri.

Are there any disadvantages to using ML-Log?
If colours are very saturated, there is a chance that one channel will clip at the lower end of the scale. I can avoid this by globally lowering saturation across the profile, but this hurts overall colour accuracy so I decided not to do that. It's probably not a big deal.

How do I install the profile for Adobe Camera Raw?
Put the DCP file here:

User Library>Application Support>Adobe>CameraRaw>CameraProfiles


What's the best workflow?
Bring in your DNG sequences to After Effects.
Set white balance in Adobe Camera Raw if necessary. It will behave a little differently to Adobe Standard. So Daylight will look different, for example. Choose either "As Shot" if you set it close to correct in camera, or tweak manually and trust your eyes.
Leave everything else at the default, including exposure. You might also like to turn off noise reduction if you will do it later with Neat Video.
Export to an intermediate codec. At the minimum 10bit 422, and preferably 444. Cineform, ProRes and DnXHD are your choices here.
When grading, simply add one of the included curves or luts. Before this curve or lut, tweak exposure. Try using a curves effect here, using only the midtones. The curve is likely to make things a little dark but it shouldn't clip too much in the shadows and highlights. If you need more or less contrast play with the white and black points as well.

Finally, I'd like to thank Danne for his support, encouragement and advice in making this. I hope it's useful to you. Here are some pics with the film curve:

Just out of interest, have you tried Koji - and if so what is your opinion of its use with ML footage?
Can you post stills direct from the timeline, please? It's a little difficult to spot subtle differebces once YouTube compression kicks in.
It's not particularly straightforward! Each film stock has different characteristics in how it renders certain hues - Portra 400 may have warm greens and tan skintones, whereas Fuji 400H will have cool, minty greens and pinkish skin, for example. In addition, contrast and saturation increase as exposure increases with film, and there are various colour casts in the shadows and highlights at different exposure levels.

We are used to exposing plus or minus one with our digital cameras, and seeing a noticeable increase or decrease in brightness. But when negative film is scanned by a lab, the scanner usually automatically compensates for these exposure differences. You end up with bracketing tests where the brightness doesn't seem to change much, but the colour and contrast does:

My methodology is first to match the general contrast and colour shift using curves, and then to fine tune exact hue and saturation reproduction for various colours using a LUT. I've dabbled with trying to introduce the film emulation as a DCP profile, but the results aren't very good. I'm getting more consistent and accurate results by debayering (to Cinelog in the case of Canon DSLR RAW but I also get great results with GH5 VLog 10bit) and then adding curves and a LUT.
Sit tight, guys. I'm gonna make a really great Ektar lut. I've been shooting lots of test charts, with lots of rolls of film, and testing lots of different ways of matching contrast and colour. Maybe in like four or five weeks I'm going to post a LUT that will match, fairly accurately, Ektar 100 as processed and scanned by one of Europe's top film labs.

Here's a taster of what I've been up to. In this case it's Portra 400 developed and scanned by my local camera shop,

and a 5D3 RAW photo to Cinelog plus curves and a LUT

This is all still a work in progress. When I'm happy I'll come back and give you guys Ektar 100... ULTIMATE!!!!