FORESIGHT.

Steven’s latest release, Haywire, brings to the screen some concepts that are relatively still “new” to digital cinema.  Haywire is a film shot anamorphic and on a 16:9 native sensor, anamorphic is a bit tricker to deal with.  Using Hawk lenses, this film has an amazing texture that differentiates it from other digital images.  These 2x Hawks actually are designed for digital cameras so they can cover a 1.33 area of a 1.77 (in this case 1.89) sensor.  The de-squeeze on the image takes advantage of the higher pixel count (in this case) of the RED MX and simultaneously allows for a reduction in overall resolution by not sampling the entire sensor.  When you see the pictures and you evaluate what makes them look the way they look, it really is coming largely from the lenses themselves.  But there are two results that are happening from this:

First, you are seeing a high resolution camera only sampling about 75% of it’s pixels.  That has an affect on the overall “precision” of the picture.  Secondly, the anamorphic glass means we have to de-squeeze the picture and stretch it from 1.33 to (essentially) 2.40.  Once we do that, we end up with a scope picture that literally takes square pixels from the sensor and makes them oblong rectangles.  The combination of less source resolution (4K to 3K) and stretching to fill a 2.40:1 aperture makes for a beautiful texture that goes well beyond the typical lens flares.  I think this makes the images more accepting of natural and enhanced vignettes and concentrates focus to the center of the frame as the edges tend to fall off fairly quickly.  A simple summation of the look of Haywire is this: optically driven texture.  People have been using optics and filtration to get looks for over a hundred years.  I always find it silly when people criticize an unfiltered digital image on modern spherical glass and say “it doesn’t look like film.”  -No kidding.

Altogether, for people that want to explore a more gentle, classic texture, Haywire is a great example and shows off a technique that is sadly not yet very popular on the big screen with digital cameras - partially due to sensor design and anamorphic lens availability.  But all that is starting to change as people see better looking digital cinema and the sensors increase in resolution and are better optimized for more textured glass like Hawks.

In the below stills, notice the vignetting, which is sometimes enhanced in DI, but it’s sourced in the lenses themselves.  

Notice the falloff of focus from the center, even on the same focus plane.  This is a nice touch to achieving a classic look with a digital camera.

The Hawks in combination with the smaller sensor area (3K) and filtration make for a very smooth textured image that looks organic, appropriately “imperfect,” colorful and smooth.

But there is another component of Haywire that I found interesting on this project and it has to do with the way in which the post workflow was evolving at the time of the DI.

In my company, we tend to favor a workflow that utilizes DPX frames as a means for optimized image manipulation.  This goes for RED RAW files as well as P2 or XDCAM media.  It is common for people to discuss the pros and cons of RAW image accessibility in DI vs. pre-debayered image accessibility, and I’ve been vocal that native support isn’t always a choice we have.  Sometimes it’s the tool that controls the workflow, other times is the client and other times its the nature of the project (a feature with 1000+ visual effects, for example).  But on Haywire, because the film was anamorphic, the director wanted to work with full resolution and still have access not only to the RAW files but to be able to do some color manipulation himself.  It was at this time that the idea of exploring RESOLVE for the MAC came to our attention and the decision was made to grade Haywire on a Macintosh.  There were two major components that made this idea very interesting:

1. We were excited about the prospect of pushing Resolve on a Mac in a large feature setting

2. We were interested in exploring the pros and cons of using native 3K anamorphic R3D files throughout the process

On Haywire, colorist Corinne Bogdanowicz used the latest release build of DaVinici Resolve for the Mac.  At the time it was still version 7, which was first major version since BlackMagic Design acquired DaVinici in early 2010.  First and foremost, Resolve has come a long way since version 7 and that’s important to this story, but on the other hand, I was excited and impressed that BlackMagic had decided to unleash this technology at a price point that made it accessible for anyone who has a computer ($1,000).  Our system included a 12-core MacPro, RedRocket cart, Decklink HD card, NVIDA QuadroFX4000 and the Resolve panel surface.  Because of the way BlackMagic does business, most people can get these components fairly quickly and fairly easily, which is a massive pro in my opinion as I am an advocate for accessibility of technology.  But the challenge of Haywire was to work at full resolution in an anamorphic setting on a 6 reel feature film.  This is where the real challenge came about that pushed Resolve and the Mac to the limits:

***Remember: EPIC and RED MX in anamorphic mode shoot an aspect ratio of 1.22:1***

1. We have a 3K anamorphic source R3D picture (2816x2304) which is roughly 6 megapixles

(this is 3x larger than standard DCI 2K and 3.5x larger than HDTV 1080p)

2. We have to debayer every frame using the RedRocket card at 24 frames per second

3. We have to desqueeze the X (east and west) every frame using the graphics card to 1.5x = 4224

4. We have to desqueeze the Y (north and south) every frame using the graphics card to .75 = 1728

(“Y” desqueezes are always 50% of the X)

This leaves us with the desqueezed resolution of 4224x1728.  That’s the equivalent aspect ratio of 2.44:1 in 4K.  So even though Haywire starts in 3K anamorphic, it ends up turning into 4K and the rendered grades, shapes, results, titles and visual effects needed to be all calculated in 4K.

5. Because the picture is 2.44, we are a little too big for cinema, so we have to crop about 1% off every frame and do a center extraction of 1% in order to make it fit academy 2.40:1

6. Then we take every debayered, desqueezed, cropped and center-extracted frame and scale it down 2048x1080 (2K projection on a Christie CP2000)

Each frame as it is played must go through every one of these steps prior to color correction.  Only after these calculations are performed can we start to color correct…

For a Mac, a medium-qualtiy graphics card, a RedRocket and a Decklink, we pushed this machine to the edge.  Literally.  The amount of calculations over the course of 9 hour coloring days was so substantial that we had to remove the side panel of the computer and blow cool air into the machine while Corinne worked in order to keep it from overheating.  There was a point in which CTO Chris Peariso actually burned himself on the NVIDIA card when inspecting it close up!  If you look at the above list, this is a significant amount of cycles that are being taken up by a tremendous amount of on-the-fly geometry and playback, thus leaving less cycles available for color and image manipulation.  This is something that owners need to consider if they’re exploring Resolve on a 4K anamorphic level with substantial VFX and plan on running it 60+ hours a week. Granted, Rev8 has included massive improvements which were unavailable to Haywire at the time, and this was the first time we used Resolve on the Mac with so many pre-grading calculations (which are not common for every project).  But given these circumstances, were we disappointed?  Of course not!  This was somewhat of a controlled experiment that Steven himself wanted to explore.  But I can say at the time we’re fairly confident Haywire was pushing a MacPro-based Resolve system to its absolute limit.  And we were collectively in favor of seeing where it could go.

Since then, Resolve version 8 has been released and further enhanced the the power that can be harnessed inside a MacPro.  BlackMagic has released a free version (Resolve Lite) and opened up control surface support to multiple systems starting as low as $1,000.  We use Resolve systems on our mobile post lab systems, OUTPOST, and I see Resolve’s used on set all over the world.  Resolve is a tool that is changing the face of image manipulation and it’s clearly picking up where Apple left off.  But what impact will Resolve have on the industry as time goes on?  Clearly there is friction amongst companies that paid $500,000 36 months ago for a system that I now run on my laptop.  Clearly there is controversy stemming from a company like BlackMagic that is transitioning into selling products that require support contracts.  And clearly there is optimism amongst a rapidly growing set of users that finally have an opportunity to practice and develop skills on the most popular coloring system in the professional world.  I believe all of these concepts ultimately point to positive things, but the industry will have to sift through a few more growing pains until this topic finally stabilizes.

Working with the Resolve on Haywire represents a great style of DI pipeline and clearly demonstrates the level of evolving creative democratization the industry is experiencing.  I want to encourage people who are exploring Resolve on all levels to keep at it because the evidence put out by BlackMagic over the past 3 years suggest they will surely keep at it, too. And while there are major benefits to the power working with R3Ds natively, there is a downside when faced with a tremendous amount of source footage in the conform (which can have an affect on GPU performance) as well as the cycles required to manage the complex arithmetic brought on by complex projects like Haywire.  It is for this reason that we still favor the Quantel Pablo system, which is capable of working with native files such as R3Ds, but we choose to pre-debayer them into uncompressed files so that the load is lessened and the computer spreads the tasks of geometry, color, scaling, debayer, decode, etc., across more processors (pic-stores) which improves overall performance, speed and reliability.  Unlike GPU-based systems like Resolve, pic-stores enable tasks to be farmed out to many dedicated parts of the machine, instead of a few.  It is also why even on Haywire why we used the Pablo for the final mastering, titles and versioning outputs to different color spaces, resolutions and aspect ratios.  It goes to show there will never be one way or “the best way” to do any task.  As the tools evolve, the best users are the ones who learn and experience the pros and cons of every system and use the right tools on the right jobs, regardless (sometimes) of how many it takes.

But in the case of this subject, the most important thing to note is that Haywire was mastered in 4K (anamorphic) and was conformed and graded on a DaVinci Resolve using not much more than standard and available tools that fit inside your off-the-shelf MacPro.  That, on every level, is significant.

| m |