All Digital Feature Film Workflow

 Blue Lies is the second film in the loosely bound Lies trilogy by doctor/director Gerald Lawson.  After shooting and editing the first film,  Little Lies in ’03-’04, I recently re-teamed with Lawson for the same double duty on Blue Lies.  In the four years between the films however, technology has moved on substantially and we were also able to take advantage of lessons learned on the first film to pursue and even more efficient workflow.  The fact that the entire post production process would be handled in-house at The Film Bakery, a production company I am part owner of, allowed even more freedom to experiment and refine the process.

Despite more than a decade of experience in managing short form projects from shoot through to delivery (even film print) it is always a shock the stop and think about the sheer scale of even a simple feature length drama.

 

THE SHOOT 

The Sony PDW-F330 camera package for the Newcastle shoot was supplied by local production company Concept and the lighting van and gaffer John McQuillan were through 180-Digital.

The F330 is a useful combination of a Sony CineAlta camera head enabling true 24 or 25p recording with cine-gammas with an XDCAM-HD back end.  The professional disc format worked extremely well, enabling us to shoot large amounts of footage very quickly with no need for on set data wrangling.  Each night I was able to plug the camera into my laptop and browse through any of the day’s shooting.  It is important to note that with the XD format, this is not the risky proposition that it is with tape bacuase the discs are true random access and very stable media.  On a couple of occasions I even cut a scene together on my MacBook to check that we were on track.

 

Another myth worth clearing up at this point is the idea that the XDCAM-HD format is just a variation of the HDV format and this could hardly be more untrue.  While both are variants of the MPEG framework HDV works with a constant bit rate of 20 mbps (actual) because the constant speed of the cassette tape does not allow for the data rate to go above this.  However, the disc based XDCAM-HD has a variable bit rate with an average of 35 mbps.  Now that may not sound like a big difference, but in practice it is massive.  The disc format allows the data rate to regularly spike far above this average and the algorithms Sony are now using seem the let it dip well below with little damage.  I have seen the data rate regularly hit over 70 mbps which is significant considering a great HD camera like the Varicam at “100 mbps” achieves that only a 60 fps.  By the time you get back to 24 fps, you are actually looking at 42 mbps, with no peaks.

What does all that mean in practice?  The short answer is that the XDCAM-HD is doesn’t just look like a pro camera, it hits well above its weight and with detailed, solid, malleable pictures and sound.

 

We shot the entire film with Cine-Gamma-4, mostly with DCC Off and all with preset white balance – mostly to 5600-K (daylight balance).  Because were had access to a good package of Kino-Flos, I chose to leave the lamps balanced to daylight and even shot night interiors with a daylight balance.  One of the great things about the F330 is that its “Virtual #85” meant that there was no exposure loss by working with a daylight balance.  In fact only one scene of the entire film was shot with a tungsten balance and this was still lit was daylight for a moonlight effect.

 

Although the F330 can record up to four channels of audio, we chose to work with 2 channels of 48k 16 bit audio.

 

THE EDIT

After completion of the 17 day principal photography we had nearly thirty XDCAM-HD discs of raw footage which equates to around 600 GB of video data.  This was transferred into Final Cut Pro via the Sony XDCAM Utility and all of the data was stored on an external firewire drive.  The contents of each disc was stored in a separate sub-folder as “Roll-01” etc. and this structure was then imported directly into the FCP project.

 

The whole project and every file associated with it was simultaneously backed up to dual SATA drives mounted into Vantec docks (which remind me of the cart systems that radio stations used to use!)  The entire cost of this triple redundancy was less than $600.

Despite a good range of metadata options on the XDCAM-HD format, we chose to go with conventional “dumb” slates, with only a traditional visual reference to Slate, Take & Scene numbers.

 

With all of the raw footage in the FCP project, the next step was to sort clips from Disc folders to Scene folders.  Another advantage of the format is that (like most data based cameras) every time I hit the record button, that became a separate clip in the FCP project.  I made sure that before I rolled on any shot, I had the clapperboard framed, exposed & focussed so that the clips could be identified from the thumbnails alone and very quickly sorted into scene order.

Initially every scene was assembled on a separate timeline within this master project and this allowed for many variations of a single scene wherever this was useful, before starting to look at the flow of the overall story.

The assembly process became a simple matter of opening the folder for the scene in question and quickly building the sequence.  The fact that I was looking through the viewfinder and already knew every shot that we had available also helped to speed the process of the first cut.

 

Although this was essentially the “offline edit” we were able to worked natively in full 1080p HD right from the beginning.  Because of the XD’s light data consumption and the rapidly dropping price of storage, it was actually more efficient and cost effective to stay at full quality all the way through the post process

 

The dialogue was assembled onto one audio track and additional tracks were used only for overlaps or essential sound effects when required.  This was mainly a house-keeping issue because adding the extra tracks in FCP is such a non-event.

The first cut was then achieved by simply exporting native XD-HD 1080p Quicktimes of each scene and then assembling them to a DVD.  The great advantage of this was that it made it very easy for us to experiment with different scene orders and even dropping scenes with very little time commitment and then distribute DVD’s to each of the stakeholders.

 

After some experimentation with the structure of the narrative we started to build the scenes together into “Reels”.  While not an absolute necessity like it is with a traditional film cut, working in reels still added immensely to the effectiveness of our workflow.  As well as giving us more manageable chunks of the project to be working with at any one time, the reels also allowed us to deal with rendering and encoding changes as well as the final colour grading and sound in manageable pieces.

 

COLOUR GRADE

Once the edit was locked off I began the colour grade.  Although Final Cut Studio and Apple Color became available during the edit of Blue Lies, we made the decision to continue with the final grade using our in-house solution, The Grading Sweet.  The great advantage for me was that we were then able to grade directly in the timeline, with titles, effect & transitions all visible but unaffected by the grade.  It also meant that we were always seeing the grade in context and with audio, a powerful factor when working with long form drama.

 

Much of this path was worked out on Little Lies and then refined while I was revamping the workflow for the Seven Network’s All Saints series.  However, on Blue Lies I was able to take this to a whole new level.  Because we were able to maintain all of the footage on the main edit system at full quality through the entire 19 months of post production, it meant that even after lock-off, the edit was always “alive” meaning that every shot was still referencing the original source material.  This meant that it was always possible to access one extra frame of a shot or an extra bit of audio when needed or even to make significant changes if they were important enough.  These decisions were always practical ones and not technical ones.

 

The film contained only one visual effects shot which solved the problem of a stunt unachievable on the film’s modest budget.  This was created in 3D-Max and After Effects by Jay Cox at Concept Creative and rendered as a Targa sequence.  This was easily converted to ProRes in Quicktime Pro and imported seamlessly into the project on its own video layer with the original shot sitting invisibly beneath it for future reference.

We also made the decision to work in Apple’s ProRes 422 (HQ) codec for the final rendering of the grade.  This gave us a very workable final data rate and a very strong picture quality when working directly from the XDCAM-HD source data.  It was pleasing to never be held back in the grade because of compression issues.

Importantly for tight delivery times, the ProRes format also produces significantly faster outputs to other formats than from the native XDCAM-HD codec.

 

SOUND

For the sound mix we made use of another “gift” in Final Cut Studio, the often overlooked Soundtrack Pro.  While there is a bit to figure out with this package, STP is a seriously heavy hitting bit of software, completely dedicated to sound for film & video.

 

We began by sending the locked off edits from FCP as XML files which we then saved as STP multitrack project files.  Although, FCP recommends a flavour of h.264 for the reference video, we found good old, boring Photo JPEG to be good quality and gave the system more headroom for handling the big project.

With all this in place we began the process of track-laying.  If you’re in the position of trying to design and mix the sound for a large project my advice is don’t skimp on track laying time… it always comes back to bite you when you do!!

From our FCP track layout we then separated the dialogue for each character in each scene onto an individual track.  This meant that compression, EQ etc. could be added to each voice to match the vocal characteristics and the environment for each scene.

A pre-mix for each scene was then done track-by-track and then all of the dialogue tracks were pushed through a Submix labelled VOICE.

ADR or Automated Dialogue Replacement, was done for a few scenes in a separate STP project before the final lock-off and these tracks were then brought into the main project for the relevant reel and treated the same as location audio from then on.

 

The original music by composer Paul Winn was delivered as separate tracks, allowing us to remix any element of the score all the way to the final mix-down.  These elements were layed into place and sent to a “MUSIC” Submix.

Atmos is such a vital part of modern sound design and Blue Lies utilised a mixture of location atmos tracks and dozens of atmos elements built into STP.  For example, the police station location required a mix of five different atmos tracks to achieve the right result.  Luckily STP makes finding and managing all of these elements incredibly easy.

Spot Effects were as essential as they are for any film with numerous phones and other momentary sounds needing to be added to the mix.  These were mostly found in the STP library and a few were recorded directly into the project from a studio mike in the suite.

 

Foley has become an art form in itself and although it is often overlooked by he audience at large, its absence is usually heard as an unidentifiable crime against humanity!

Although STP makes the addition of Foley effects for footsteps, rustling clothes and other rich sound we take for granted as easy as it can be, finding and synching these effects is still a great challenge for low-budget sound design.

 

Compared to the  Little Lies mix only three years earlier, we were able to do a massive amount of Foley work on Blue Lies.  Having said that, this is probably the one area where I wish we had had another six weeks of tweaking-time..!!

The spot effects and Foley were both pre-mixed track-by-track and sent to a Submix labelled “FX”.

 

The final mix of each reel then began with four Submixes –  VOICE, MUSIC, ATMOS and FX.

We made use of a wonderful little mixing panel from Berhinger called the BCF-2000.  With 8 motorised, flying faders, lots of buttons and 8 pan potts, for around $400 this USB gadget simply made the process possible.  One key element of this was adding the LC-XMU software (very cheap) which provides an onscreen display which relates the BCF-2000 faders to the STP track names but also allows you to re-program functions of the mixer for full touch-style mixing.

For the final mixdown of each reel, we made all of the pre-mixed audio tracks invisible in STP and dealt only with the four sub-mixes, although the premixed tracks were all still available for the occasional tweak.

These mix-downs were then exported as AIFF files and mated to the vision back in FCP.

 

For the DVD we exported all of the reels to  Quicktimes (just under 100 GB total) and encoded in Cleaner before mastering in DVD Studio Pro.

The next challenge was preparing the film for a digital cinema release.  This is currently a technical and fiscal possibility like never before.  But if you can find any useful specifications for digital cinema formats you are likely wasting you life as a film-maker and should be a detective.

Luckily we were able to work closely with one patient and cooperative independent cinema to develop an encoding profile that has worked remarkably well.

 

When we finally delivered the E-Cinema version to the first cinema with a full 48 hrs to spare, the results were no less than thrilling on the big screen.  And it was no small satisfaction that the entire process had been achieved with one camera, a standard MacBook, an old (but reliable) G5 and a bunch of hard drives.

 

Exciting times to be a film-maker.

Comments are closed.