In-camera grading. Good idea or not?

[S.Stefan]

I'm new to video so bear with me. I've shot stills for a long time though, both professionally and for my own pleasure.

 

Now that I'm educating myself in the various ways of the video world I find myself pondering a question regarding tonal and color editing, i.e. grading.

 

I'm currently shooting micro 4/3 equipment, specifically the Olympus OM-D E-M5. I will be getting the E-M1 soon. While these cameras are generally not highly considered in terms of video specs they are fantastic still cameras and I am getting perfectly good results shooting Motion JPEG's and editing short sequences in Photoshop. This codec seems reasonably easy on my computer's CPU and seems to allow for some editing leeway. Unfortunately the camera only has a 30fps option which is not ideal for me as I live in a 25fps PAL area. Very impressive in-camera stabilization though. 

 

Anyway, it seems to me that since video is highly compressed 8bit it tends to break apart with heavy editing. Getting it as close to perfect as possible in the camera would seem to be a good idea. Bearing that in mind the live curves tool built into both cameras should be useful tool for shooting video. Using that in conjunction with the live on-screen histogram available in both cameras and the new Color Creator feature in the new E-M1 camera for custom tailoring individual colors should allow one to get pretty close to a final look in the camera.

 

If I'm right this could both be time saving and serve to maintain image quality in post.

 

Is my reasoning on this correct or am I missing something? I am after all new to video.

[Zach Ashcraft]

You might get a variety of answers here, but for my line of work I do try to get it as right in camera as possible. I primarily shoot weddings, and it is a huge time saver to not have to grade footage after the fact. Of course there are almost always slight exposure or contrast adjustments for me, but for weddings I never shoot a super flat image. 

 

If you're shooting a short film, many folks want to shoot a flatter image so that thy have more creative control in post. However you are correct in that editing the codec of that camera will likely degrade the image. It'd be good to shoot some footage with it, spend some time trying to grade it and figure out the limitations of the footage

[Axel]

Anyway, it seems to me that since video is highly compressed 8bit it tends to break apart with heavy editing.

 

 

Not necessarily. I admit, there are limits, and that's why nowadays raw is on everyone's lips. But that's not because a codec 'breaks apart', it's because you have a lower range of values. It is important to understand the difference, especially since you are asking for an intelligent compromise.

 

I always color correct my footage, and I have been doing this in 2002 already, when I knew nothin about 'grading'. The best shot can nonetheless gain from a subtle polishing, always. 

 

In the blog of famous DOP Shane Hurlbut, he admits to be an old school type of photographer, who tries to get everything right in-camera. What does it mean? He thoroughly lights the set, he carefully measures exposure and he sets the white balance manually to fit to the mood he tries to achieve.

 

The last step introduces a problem. A white balance that is not set to white will shift all values in the image and leave a cast, similar to that of a color filter (though of course a color filter brightens it's own color and darkens the opposite). With secondary color correction, one can do much more subtle changes, and, for example, protect the skin tones, which in most cases will result in a better image. You can apply a correction to every color separately, stack secondaries upon each other - but only if you start with a perfectly balanced, evenly exposed image in the first place (both the goals of primary cc).

 

And if you do these extreme changes, will the 8-bit, heavily compressed stuff fall apart?

 

This depends on these things:

 

1. The accuracy of the application you grade in. It has to be 32-bit floating point computing, which means there will be no banding or other artifacts caused by your new mapping of tones.

 

2. The codec and it's implementation. HDV (or XDCAM EX) often can't even be brightened without reaching it's limits. Mpeg4 of the EOSs or the GHs usually works well (the latter particularly with higher bitrates).

 

3. If your changes stay within the range you recorded. This is the important point. This is what you need to be aware of when shooting. It's also true for less compressed video like 10-bit ProRes or raw, only those are more forgiving. This is what you can do in-camera. 

 

Is post a pita? I don't think so. It's creative fun. But that's a matter of personal preferences.

[jgharding]

A bit of both, is a short answer.

 

Super flat compressed footage with quantizd 8-bit colour values can't be pushed as far as you like, but if your in-camera look is too strong you're committed.

 

Get white balance correct in camera either way. Perhaps use a slightly flat profile, just to stop black getting crushed, and give yourself some room for movement.

 

It's certainly useful to decide your outcome look before you start, if you can.

 

The 8-bit banding can be helped by adding grain in post. This acts a bit like dithering.

[S.Stefan]

Thank you both for replying, Zach and Axel,

 

Axel, I totally agree with the creative importance of post. I do not consider it a pita. I've been shooting raw for stills for over a decade and love the flexibility it offers for tonal/color editing. I do not have a raw capable video camera yet and although I'm keeping an eye on the Blackmagic Pocket Cinema Camera I probably will not be shooting raw video in the near future. I expect it will be too taxing on my current computer equipment. Possibly later.

 

In the meantime my thinking was it might be a good idea to come pretty close to a final image in-camera in order to minimize chances of banding etc. in post due to possible heavy editing. I realize some editing will in my mind always be necessary to fine tune the look one wants but if one can do that with reasonable confidence that the image will not degrade and the light touch needed will not be very time consuming it should be worth the in-camera effort.

 

Being new to video I have come across recommendations on the internet to just shoot with a very flat profile and fix in post. Just sticking an arbitrary flat profile in there and fixing in post didn't seem to make perfect sense to me if I have a live curves tool in the camera where I can adjust highlights and shadows separately.    

[S.Stefan]

Thanks for your reply jgharding.

 

You were a couple of minutes ahead of my last post.

[S.Stefan]

A bit of both, is a short answer.

 

Super flat compressed footage with quantizd 8-bit colour values can't be pushed as far as you like, but if your in-camera look is too strong you're committed.

 

Get white balance correct in camera either way. Perhaps use a slightly flat profile, just to stop black getting crushed, and give yourself some room for movement.

 

It's certainly useful to decide your outcome look before you start, if you can.

 

The 8-bit banding can be helped by adding grain in post. This acts a bit like dithering.

Thanks. This makes sense.

[jgharding]

Indeed raw digital photography is the ideal, and hopefully within a couple of years video will behave in exactly the same way ad computers will handle it with aplomb!

 

At the moment video is a compromise, but great results can still be achieved.

 

Consider if you were shooting still JPEG instead of raw: You'd want to get a decent looking result but leave a little wiggle rooms for adjustment, I think H264 video is a fairly similar way of shooting!

[jgharding]

For example I use a profile called VisionColor for Canon EOS. It emulates Kodak tones subtly, lifts the blacks and flattens things a little, but still hold the look of the scene. I find it a good compromise, while my tests of ultraflats often show slightly lifeless results...

[S.Stefan]

...

Consider if you were shooting still JPEG instead of raw: You'd want to get a decent looking result but leave a little wiggle rooms for adjustment, I think H264 video is a fairly similar way of shooting!

 

Thanks, does Motion JPEG have more wiggle room than H.264?

[jgharding]

I don't know much about that Olympus I'm afraid! Someone here may though... it'll depend on how it's implemented I think

 

All i know is that the stabilisation would have sold me one, if it hadn't been for 30p. Such a silly oversight from Olympus there.

[S.Stefan]

I don't know much about that Olympus I'm afraid! Someone here may though... it'll depend on how it's implemented I think

 

 

OK, thanks. I hadn't thought of possible different implementations within the codec. I'm pretty familiar with JPEG stills and if MotionJPEG is just a series of JPEG stills I think I have a fairly good idea how much editing leeway I have. Not familiar with how much editing H.264 will take. 

[jgharding]

it's a per camera thing. Test it and see for yourself, I'd say!

 

try upscaling the 720p MJPEG next to the 1080p AVCHD and see which you prefer working with....

 

One can easily get abstract in terms of the algorithms involved, but suffice it to say that MJPEG is less complex, it
takes less power to encode and decode, and is less efficient. In theory it can provide better quality as it's I-frame (it stores each image individually, doesn't use big groups of frames and prediction), but if you hack a Canon EOS you can have I-frame adaptive-high-bitrate H264. But the sensor downscaling is soft. So again it depends on the camera, not the stats or acronyms.