As promised, we'd like to bring you some of the ARRI Alexa test footage images that we captured and worked with in the DI theater.
Overall, we were extremely impressed with the Alexa. The images came together so quickly and easily, with beautiful results. For a DI color facility like us, we specialize in making the final movie look the best it can. Here the Alexa image makes it a lot easier to get there, which is a more enjoyable experience for us. Instead of worrying about how to prevent skin-tones from looking plastic, we can focus in on the emotional resonance of our grading choices.
We used the Alexa's DTE "direct to edit" functionality, which records to Quicktime ProRes format right on board to SxS cards. We opted for the higher quality capture, which is ProRes 4444, a 12-bit RGB codec, and recorded in LogC colorspace (more on that later).
To kick it off, here is one of our favorite daylight shots, out by the ocean in Santa Monica. This is without any reflectors or bounce boards. We did use ND filters in front of the lens, in order to cut the light down for the sensor, which is rated ISO 800. Click on the image to get the full resolution JPEG, or on the links beneath for the uncompressed 16-bit TIFF files.
Now, here is the image in LogC color space. This was captured straight from the Quicktime ProRes 4444. Again, click on the image for the full-res JPEG or below for the TIFF.
Most involved in the post business are already familiar with "Log" (logarithmic) as a method for digitally storing image data. For those unsure, a brief explanation: Log refers to the brightness curve applied when transferring the image data to the digital bits of data. Using a Log curve is better able to mathmatically represent the image data in the highlights and shadows where the human eye generally ignores those details. The mid-tones are unaffected, because they still have plenty of data representing them as well. But when we look at it with the human eye, it looks flat and grey-- no contrast. Generally we're used to seeing "Linear" images, which have a gamma curve applied to the image data to better represent how we see with our eyes. The mid-tones are over-represented, and the shadows and highlights are crushed.
"LogC" as ARRI calls it, is really a designation that they came up with to describe the logarithmic curve that they're applying to their image data before capturing or compressing it into the image data stream. In order to look at the image properly, it will need a color grade or "look up table" transformation to adapt it to linear space.
The bottom line is that Log data gives you more flexibility in the color grading process. You have tons of data in the shadows and highlights so that if your image is underexposed or overexposed, you can adjust the grade to still make the final image look great without gathering too much grain or banding.
Here's a Histogram of the LogC image. As you can see, most of the data is bunched up in the middle, with tons of space on either end to represent highlights and shadows. Again, this doesn't mean the mid-tones get short changed, they still have all the data bits they'll ever need.
As you can see, we're pretty much not clipping anything. We didn't have to choose between exposing for the shadows or exposing for the sky.
Here is a few seconds of the actual Quicktime ProRes 4444 footage, so you can play with it yourself. Right-click HERE to download the .MOV file. It's big (258 MB).
Now on to another daylight challenge. In this image, Rain's face was completely in shadow against the bright sunny sky. Here's the LogC (click for TIFF).
And here is the grade.
Here is another shade shot, this one with a lot of detailed trees & leaves to try and push the compression scheme.
Now here is a shot from our night footage. We didn't plan anything, we simply took the camera outside our front door and used only available street light. This was rated at ISO 1600, and we didn't have Rain standing under any street lamp or direct light source. There is a green cast on her that is coming from a small decorative light on our building. Here we see what the grain looks like, but definitly still end up with a very usable image. As far as our night tests, we discovered the Alexa was literally more sensitive to light than the human eye.
And for those interested, here is a few seconds of the full ProRes 4444 LogC video file. Right-click HERE to download the .mov (134 MB).
Last, here is the entire graded video, web compressed to H264. Visit Vimeo to download the full 1920x1080 MP4 file.
Because the Alexa has such a high dynamic range (HDR), when we're in LogC space we capture both the details in the shadows as well as details in the clouds and sky. Often the challenge in the DI process is to try and keep as much of that information as possible when moving to a more "human eye" grade in HD or cinema color space. To grade favoring detail in the midtones is to loose detail in the highlights and shadows.
This is a good problem to have. With all the grading tools available to us there are often tricks we can use to keep the detail in a blown-out window in a dark interior shot. Also something we'll be exploring is motion tone-mapping, which is a technique that still pictures have been using for a long time (Do a quick search for HDR and tone mapping to see the results of that).
Next week, look for a discussion of your workflow options with the ARRI Alexa. Shooting in LogC presents special challenges for dailies and editorial. Then in November we'll see a roll-out of the ARRIRAW format, which gives even more flexibility (and challenges) in post.