Topics

Hello all,

I'm not sure this post fits here cause it deals with other aspects apart from post process, so if not please feel free to move it.

This is a short test I did in order to learn how to properly expose when using ML Raw on 5DmkIII. Apparently in order to get a cinematic image with the 5DmkIII the exposure is EXTREMELY important. Well exposure is always super important but it turns out that with 5DmkIII exposing in a certain way can lead to either a very cinematic or very video-ish image. I don't have any scientific explanation for this, but in my testings I found that underexposure is the key to a cinematic image with the 5DmkIII. Again I stress in MY testings. This works for ME and the way I shoot. This might not work at all for you.

Here's what I did:

I shot 5 clips. One at correct exposure and the others where either over or under exposed. Here are the exposures:

1. f/4 (this one is overexposed by one stop)
2. f/5.6 (this is the correct exposure according to incident light reading)
3. f/8 (well obviously this is underexposed by one stop)
4. f/11 (well two stops underexposure)
5. f/7.1 (underexposed by 2/3 of a stop)

The shutter was at 1/50 and this was shot at 25fps, 1920x1080 ML Raw.

Workflow:

1. Converted from Raw to DNG using mlrawviewer.
2. Opened DNGs in Davinci Resolve lite 11
3. Applied the same LUT for all clips
4. Adjusted exposure on each clip (except for f/5.6 which remained untouched) to match the exposure amount on models face, specifically forehead in the f/5.6 clip. I adjusted the exposure mainly by changing the exposure value in Camera Raw tab but also by adjusting the shadows and highlights.
5. Development setting for DNG where: Decode: Clip, Color Space: BMD Film, Gamma: BMD Film 4K, White balance was As Shot which was 5400K. I checked "Highlight Recovery".
6. Exported as ProRes 422 stitched together in Premiere Pro and uploaded to Vimeo.

To me it looks that the proper way to expose 5DmkIII for cinematic image is to expose for highlights (to the right) and than underexpose by about 1.5-2 stops depending on the scene contrast. In post you must increase exposure and correct highlights and shadows. I increased the exposure by about 1.5. Highlights were increased by about 30 and shadows increased by about 10. Without increasing highlights it looks a bit lacking contrast. This changes everything in my opinion. The highlights look nicer, they roll-off nicely and the same goes for the shadows. This puts the middle grey at a more balanced place. Overall the image looks more organic and pleasing. The colors look less "Canon" this way and more "cinema". Of course this may introduce noise, but in my testing you're good to go up to about 1600 ISO. Maybe I'll publish my other test at some point.

I'd like to hear what do you think of it? Recently I've been able to intercut the 5DmkIII with Alexa on a commercial shoot (just for kicks) and they played nicely together.

Hello,

Is there a way to play raw video with external monitor connected?
I'm using 5dmkiii 1.2.3 with latest nightly build and if i press play with monitor connected i get "file not found" error. Without monitor, the file plays just fine.

Thank you,
Goran

Hello all,

I'm using 5dmkiii 1.2.3 with the latest nightly build (29th March 2015). I upgraded yesterday.
After taking some test shots, I turned the camera off and took out the CF to download the footage. After putting the CF back into the camera and turning it back on I got a message along these lines (cannot remember the exact message): "Disabling modules load, the camera didn't shut down properly/Camera shut down ugly, disabling modules loading"
I turned the camera off and back on and everything is working fine, I was just wondering what's going on and what has happened? This is the first time I'm seeing this message in more than a year of ML usage and nothing has changed in a way I'm turning my camera on or off

Thank you,
Goran

Hello All,

Here's a short test I did:

This is a short and not very scientific test I did exporting the footage from Mlrawviewer in different log spaces to see how well they would look and grade and to see how wide of a range of usable Log spaces I could get from the 5D mark III footage.
All footage was shot with 5D Mark III in raw with a Canon 50mm f/1.4 lens at 1920x1080 resolution. Footage was opened in Mlrawviewer 1.3.4 on a Mac and exported in four different log spaces in this order: C-Log, Log-C, S-Log, S-Log2.
The files were exported as ProRes 4444. In Davinci Resolve Lite 11 I applied different LUTs from the excellent VisionColor Impulz package which were matched for the type of footage. For example, the C-Log footage had a C-Log version LUT of the Impulz emulsion, Log-C had a Log-C version of the LUT etc. Every shot received the same LUT and I have not changed the footage in any way except for applying a LUT. The LUTs have changed between different shots but for the same shot I used the same LUT matching the Log space of the clip.
I cannot remember the exact LUT I applied but it doesn't matter that much as I made sure I'm applying the same LUT for every clip in order to be able to compare the footage. The only thing changed, as I mentioned, is I matched the type of LUT to type of Log footage.
Some additional details:
First shot was shot under tungsten light at 800 ISO at f/1.4.
Second to fourth shots were shot at 100 ISO at f-stops between 16 and 22.
The last two shots were shot using a daylight coming from behind the camera at ISO 400 with f-stop around 4 but I'm not sure about that one.
Apart from Log-C not really being suitable for the 5D footage except for only the brightest of conditions I think all other log spaces are usable, with C-Log being clearly the best match for this camera.

I'd love to hear your opinion.

Goran

Hello all,

This is a fashion film I've done for the Laisha Magazine in Israel. It complements a 10 page editorial in the magazine.

The H.264 compression on the background is ridiculous, but the source material is gorgeous with nice gradients.

The video:

The editorial isn't published yet. This is the only image from the editorial that has been released so far:


It was shot on 5dmkiii using .raw.

If you have any questions, feel free to ask.

All best,
Goran

Hello,

Here's a quick question. Let's say I load the following modules:

mlv_rec
mlv_play
mlv_snd
file_man

If I never play files on the camera I can disable the mlv_play from loading. However for the one time I might use it, is it ok to leave it loaded? Does loading several modules even though not in use has negative impact on performance? If I load only raw_rec for example will camera have more free memory and thus "perform faster/better"?

Thank you,
Goran

Hello,

I was doing some testing with the 4th of July build with external monitor and mirroring on.
One thing I noticed is that the ML overlays do not cover the entire shooting area on external monitor.

Here's an image that will explain what I mean:



Notice how the ML overlays cover about 85% percent of the actual shot area. The area that gets recorded is the entire area visible on the screen including the areas outside the ML overlays. On 5D internal display they of course look good and fill the entire screen. Do not confuse it with the fact that the entire image fed from 5d doesn't fill the entire monitor screen. I'm not talking about that. I'm talking about ML overlays covering less than the entire image fed from 5d to monitor.

5D MK III 1.2.3
Resolution 1920x1080
Raw

The question is, why? A Bug or once again some setting I missed?

All best,
Goran

[Without]

Hello,

I thought I might report something I encountered today.

I was shooting with 5D mark III 1.2.3, latest daily build from 21 of June with .mlv Raw. 64gb Komputerbay x1000 cards.
I used 2 external monitors - one for myself, the other for the focus puller. The first monitor connects to the camera and outputs the signal to second monitor.

When shooting with the monitors after about 30 sec. a frame would skip and recording would stop. Without using monitors it doesn't happen! I can record until the card is full! I tried also the nightly from 23 of May and it behaved the same. With monitors max 30-ish sec. of recording before skipping a frame. Without no problems what so ever.

When I switched to .raw instead of .mlv I could shoot without any problems with both monitors connected.

My question is:

1. Why? What might have been wrong? What might be the cause to the problem?
2. What can I do to help solve this problem? What info/testing would you need me to do to so I can be of help?

All best,
Goran

Hello,

I couldn't find and answer for what seems to me like a basic issue so I'm posting here. Sorry if it's been answered before, I really couldn't find an answer.

In order to go up from one nightly to the latest nightly build I simply replace the ML folder, autoexec.bin and ML5D3123.FIR on the card. I guess that's the right way to do it. However, this looses the customizations I've made such as resolution, loaded modules, etc. Is there a way to upgrade without losing the customizations?

All best,
Goran

Hello,

There are three builds from the 20th of June, all built within a range of about 30 min. with only the first one (oldest) having a changelog. Two newer builds don't have any notes inside the changelog so I'm wondering if these are idenitcal or is there an actual change between the last three?

Thank you,
Goran

[Camera Raw]

Hello all,

After searching online for the best way to process ML Raw files for use with VisionColor Osiris LUTs and not finding a clear answer, I decided to do some testing on my own and here's the process I've found to work the best for me so far.

The main problem with using VisionColor Osiris LUTs is that they're really "optimized" to work with their VisionLOG Raw and that means using AE for handling the Raw files which is slow. If you're using Davinci Resolve to handle your CinemaDNGs you're stuck with the BMD Film for debayering and gamma which is, well, made for completely different camera than the 5D MK III or any other Canon and applying the Osiris LUTs to BMD Film processed files gives a completely different colors from the ones you get using the VisionLOG.

Of course you could match the colors from BMD Film processed CinemaDNGs to the ones processed with AE and VisionLOG but this workflow is for the "one-click" type of work where you get 95% of the way with just applying the LUTs.

So here's my workflow, I hope you find it helpful:

1. Convert .mlv files using RAWMagic. Check 16 bit CinemaDNG Output. Check Vertical Stripes Correction.
2. Import CinemaDNG files into Davinci Resolve.
3. Debayering in Camera Raw:
   Decode: Using Clip
   White Balance: Choose Appropriate Value
   Color Space: BMD Film
   Gamma: BMD Film
   Check Highlight Recovery
4. Create two serial nodes.
5. On the first node apply BMD Film to VisionLOG LUT 64 version. You may download the LUT here: http://www.vision-color.com/visionlog/
6. On the second node apply creative LUT such as VisionColor Osiris M31 for example or any other Osiris LUT.
7. Adjust exposure on the first node using Camera Raw accordingly. I've found that you may need to adjust the exposure up to -2 on properly exposed files (incident light meter reading) in Camera Raw since the BMD Film to VisionLOG LUT will make the image much much brighter. I don't think it's a good exercise to underexpose during filming because it may introduce grain. At least in my testing properly exposed files at 800 ISO look clearer than the underexposed ones by one stop at 400 ISO. Mainly in the shadows. From what I've seen the BMD Film to VisionLOG LUT doesn't blow any details on properly exposed files, so decreasing exposure in Camera Raw retains all of the original info in the file. Underexposed files show excessive grain after applying BMD Film to VisionLOG LUT.
8. Make any additional adjustments.

I need to test this a bit further but so far that seems the best way to me.

I'd love to hear what you think of the process and I hope you find it useful.

All best,
Goran

Hello all,

I created a Fincher ground glass crop marks for Magic Lantern for you to use. First here's a download link:

https://dl.dropboxusercontent.com/u/40313605/Common%20Top/CmmnTop.bmp.zip

For those of you not familiar with the Fincher crop marks here's some background even though I'm pretty sure most of you know this stuff.

There are many aspect ratios that are used in cinema. Most common ones are:

4:3 - The academic ration. Old square-ish ratio.
16:9 - Most common ratio currently among TV materials and of course web.
1:1,85 - One of the two cinematic ratios. Usually associated with the lower budget non-Hollywood movies and arty stuff. The main reason why it was used is because it's wider than the the 3:4, but used to be cheaper comparing to the anamorphic process when shooting film. Not sure if it's really used today cause it's almost identical to 16:9.
1:2,40 - The modern widescreen. Long time ago it was 2,35 and now days sometimes it's still referred to as 2,35 though it's technically a mistake. Previously expensive and associated with hi-budget only as it was used with expensive anamorphic lenses and required much more light, because the lenses had to be stopped down to around f/5.6 in order to get a sharp looking image. That resulted in a much higher production value and was accessible almost exclusively to expensive Hollywood movies.

Today there's no financial aspects involved when choosing a format so choosing a format is either strict (like when shooting for TV) or totally a matter of creative needs.

One thing you should consider is the TV factor even when you shoot for the big screen as almost no tv channel will broadcast a widescreen picture and they will crop it to 16:9. This can be done either in a simple way - just cropping equally from the left and right sides, or with a pan-scan, meaning choosing crop for every shot and sometimes adjusting the crop dynamically during the shot in order to not loose important objects. The main problem is that you don't have control over it and your film might end up awfully cropped when broadcasted on tv.

The most common way to deal with it is framing your 2,40 with a 16:9 "safe zone" inside of it - meaning not including important things outside the central 16:9 zone of your frame, so when it's cropped by the tv channel - everything will still be more or less in place.

Personally I don't like that idea cause it limits your framing freedom and you're limited in fully using your widescreen frame. So that's where Fincher glass comes into picture. This is something David Fincher came up with quite recently. In Film cameras where you have optical viewfinder there are replaceable ground glasses with different formats (just for curiosity take a look here, but these are not all available: http://bit.ly/1xzV5qp ). One is called "common top" that means that the 2,40 area is not in the middle of the frame but in the very top leaving extra space below it. Fincher realised that a much better way is taking your 16:9 sensor area and putting the 2,40 area not in the top or middle, but a bit lower than the top. Leaving 1/5 of the extra space at the top and 4/5 below.

Here's how it looks:


That way you can frame your 2,40, just making sure you don't include any unnecessary stuff in the whole 16:9 image and later you can export 2 versions of your footage - one with the 2,40 crop and the other one - full frame 16:9 and both will look good cause in most cases adding space below your frame won't kill your composition and to compensate the relatively big addition to the bottom, you add a little bit (1/5) on the top. That proved to work really good!
There's also a version with 1/4 offset instead of 1/5. The one I made is the 1/5.

The Common Top crop marks I made include the 1:2,4 crop marks inside the 16:9 aspect ratio as well as the classic 4:3 crop marks on the sides. I hope you'll find it useful the way I'm finding it useful.

How to "install":
Just copy the .bmp image inside the cropmks folder. The image is already rle compressed.

How to use:
Set your camera to 16:9 aspect ratio and load the Common top crop marks and frame your movie using the 1:2,4 marks making sure nothing important is located below and above the marks. I'm using these crop marks with 1920x1080 resolution shooting in .mlv raw and it works fine.

Since ML is such a gift to me, I'm trying to give back something and that's why I made these.

A huge thanks goes to my DP friend Mark Ziselson who pointed me in this direction. You really should check his last two features shot in 1:2,4 using this very technique.

https://vimeo.com/96342836

https://vimeo.com/77467584

These I'm afraid where shot with Alexa though

All best,
Goran Ljubuncic
www.goranphoto.com
https://vimeo.com/goranphoto

Hello,

After searching this forum extensively, I've yet to find an answer to this so I'm asking for your help.
I'm using the latest nightly with 5dmkiii 1.2.3, recording in mlv with sound.
In the menu I set aspect ratio of 1:2.35. Before I hit record in Live View I see the image in the correct aspect ratio. Once I hit record, the whole screen fills up with the video and the aspect ratio of what I see changes. This way I cannot know what is in and what is outside of my frame.
The file gets recorded in the correct 1:2.35 ratio.

Is there some setting to change this cause I haven't found or is this a bug?

Thank you in advance,
Goran

Hello,

Just installed ML after being long time on the fence. Great work! I finally get quality I want for my videos.
2 issues I noticed:

1. When I press the SET button to zoom x5 and than x10 to check focus, at x5 zoom the image is pixelated, blurry and black and white. At x10 zoom it looks ok just as always and I can check the focus. The problem is, sometimes it's easier to find focus at x5 zoom instead of x10 zoom so it would be nice for it to work properly or am I missing something?

2. The Live View image of the video doesn't fill the entire LCD on the camera back but instead it covers about 80% of the screen. I'm attaching an image to show what I mean. Once again, am I missing something is this a feature, a bug or a changeable setting?



I'm using the latest nightly 1.2.3 firmware build fro 5dmkiii.

Thank you,
Goran