Thursday, May 10, 2012

The Old Man and the Taily-po - Production Notes

So, around the end of March, I finished my first real short film since film school. It's called "The Old Man and the Taily-po", and it's based on a folk legend with a ton of variations around the deep south, Ozarks, etc. that's usually told as just "Tailypo". I adapted my version to the area of New Mexico that I live in. Here's the short itself:



If you'd like to read more about the legend, here's the Wikipedia entry: http://en.wikipedia.org/wiki/Tailypo

The short was primarily meant to be an entry into YouTube's "Your Film Festival", but it was also an exercise for me in how much of this story could be done with (basically) no budget.

My production equipment was:

  • Canon T3i/600D DSLR camera
  • Canon EF-S 18-55mm f/3.5-f/5 kit lens
  • Canon EF 50mm f/1.4 prime lens
  • Zoom H4n digital audio recorder
I considered renting a better lens, but two decisions changed my mind:

  1. This project was going straight to YouTube, where overall composition and lighting would make more of an impact than raw sharpness.
  2. I was seriously considering converting the whole project in black and white, to give it a "Twilight Zone" look.
It turned out that I would have been hampered by using a bunch of other lenses anyways, because I shot fast, and even changing between the two lenses I had was a luxury.

The actor, dog, locations, and music were all obtained either for free, trade, or very reasonably. This has been a great upside of the area I live in at the moment: There are a lot of people around here with passion about art, music and movies who love to just go out and do the things they love, and are also experienced enough to set proper boundaries about what they will and won't do.

I shot over the course of four days: two at outdoor locations and two at a cabin built as a movie set. I was wildly ambitious about how much I could get done at the cabin set, but it was also a great experience in shooting incredibly quickly.

How quickly? Try 30-40 setups per day at the cabin. The outdoor shooting was much less ambitious, but I also wasn't under the same time pressure to do so.

This speed was only possible because of the help of several other people, a surprisingly cooperative (untrained) dog, and the versatility of the T3i: I didn't have a ton of equipment jamming up the set, and I could very quickly re-position the camera to get new and interesting angles.

I shot about 95% of the footage using the Similaar "Flaat_2" picture profile. Because I was going to export to Youtube, the very slight in-camera sharpening would be acceptable, and the shadows wouldn't have to be wrestled back into place like they would with Technicolor's Cinestyle profile. I could have also used the Marvels Cine V3.4 profile, but (I reasoned at the time) in the event that I decided to go with color footage, Flaat_2 has better default skintones (a somewhat reddish tint, same as the "Portrait" picture profile on which Flaat_2 is based).

All three profiles mentioned above are designed to create a more "flat" image that decreases the amount of contrast so that more detail information can be recorded by the camera in the shadows and highlights without a "baked-in" look that cannot be adjusted significantly in post. It's kind of a kludgey way to imitate the "raw" mode of higher-end digital cinema cameras, but I've come to appreciate using it since I don't carry around a portable video monitor with false color/vector scopes/etc.

Anyways, the point of choosing a flat picture profile is that it gave me enough image information to simply shoot by exposing what I wanted to show; in scenes lit by practical sources. This did not mean I  ignored lighting, just that I didn't have to add a significant amount of light to most scenes. It meant that I could use a 75-watt incandescent bulb to light much of the cabin interior shots at night, and that was enough light for my particular aesthetic - which was to only use motivated light sources. For some shots, I only used a Coleman lantern. For this story, I wish I could have used it more.

There were some downsides to this approach. For many nighttime cabin shots, I had to shoot at around f/1.4 to keep my ISO down. Shallow depth-of-field junkies love this, but if you're trying to keep an actor's face in focus, it's a pain in the ass. Since I didn't have a follow focus rig, I decided to just let the shots go out of focus when my actor moved a fraction of an inch. Also, it limited me to the 50mm lens, - great for close-ups, not so much for wide shots. For the most part, I don't think it turned out too badly, but in the future, I would love to use a Canon 5DmkIII and just increase the ISO.

Speaking of high ISOs, one of the expenses on this short was buying a copy of the Neat Video plugin for After Effects. It's not a click-and-forget sort of plugin - you do have to adjust and fine-tune every shot you use it on - but it allowed me to take 3200 ISO shots and make them look pretty darn clean. I even used some 6400 ISO shots that were lit only with two tactical flashlights; I wouldn't recommend doing this due to the awful amount of noise in the image, but as long as you keep the shots short, Neat Video can make a nasty-looking image look significantly more acceptable.

I'm currently still using the Adobe Production Premium CS3 Suite, which is getting very long in the tooth. At first, I wasn't even sure I could do a real workflow for H.264 material without transcoding everything to an intermediate codec. Initially, that intermediate codec was going to be Cineform, but my current hardware/software combination made that unrealistic, so I figured out an offline/online edit workflow.

My post workflow was:

  • I transcoded 23.976fps 16x9 anamorphic DV25 (aka "MiniDV") Quicktime proxies from the original camera footage with the exact same file names (obviously rendering them out to a different directory than the originals). I used MPEG Streamclip for the transcoding, and I highly recommend it for batch-encoding to Quicktime formats, especially since it will preserve the original framerate settings by default.
  • I edited the short in Premiere Pro CS3 using the DV25 proxy files.
  • When done, I exported a mixdown of the audio tracks to a single stereo .WAV file, then imported it into the project, put it into the timeline, and muted every other audio track.
  • I opened up a new uncompressed 1080p/23.976fps Premiere Pro project and imported the DV project file. I made all the video clips offline, then re-linked the clips to the original footage. Without opening any of the sequences, I saved the new project and quit. If I tried to open/scrub through any of the sequences, Premiere Pro would run out of memory and crash.
  • Opening up After Effects, I imported the uncompressed Premiere Pro project file. I went through the main sequence and checked to make sure that all the clips were properly linked to the original footage, at the right size, and I removed all the audio tracks aside from the mixdown track. 
  • I then went through and did all my color grading clip by clip using Magic Bullet Looks, and noise reduction using Neat Video.
  • Once everything looked good, I rendered out a Blu-Ray, DVD and Cineform Quicktime copy. The Cineform version was used to encode to an internet-friendly H.264 version for Youtube using MPEG Streamclip.
And some things I learned:

  • I enjoyed directing and shooting the short, and I will likely direct, shoot, etc, again.
  • The media file swapping "trick" only works if you keep the framerate the same between the originals and the proxies, so I don't believe 720p/60 footage could be edited accurately using DV25 footage (it only does up to 29.97fps). 
  • However, I did end up using 30p and 60p footage that was played back at 23.976fps for some slow motion shots. You can set this via the "interpret footage" menu option in Premiere Pro or After Effects. I believe I had to set it in After Effects, so it's not ideal for getting the timing precise during the DV25 proxy editing for the above-mentioned reasons. I adjusted the cut in After Effects but kept it the same length so audio sync would be maintained. Yeah, I know, kludgey; but it worked.
  • For greater precision with editing slo-motion clips, you could import the higher-framerate clip into After Effects on it's own, set the playback speed to your project framerate, and render out to an intermediate format (like Cineform), which could then be used to generate a proxy. This way, you wouldn't have to adjust anything later.
  • You need to set your memory limits properly in AE, or AE will crash constantly. I used the /3GB memory switch to allow Windows XP to give AE more memory, which helps if you have more than 2GB of RAM installed.
  • Only basic dissolves, image size/motion and framerate changes will survive the After Effects import process.
  • You need to apply Neat Video before Magic Bullet Looks (or any other effect) in order for the former to work.
  • I ended up not having enough hard drive space to transcode all my footage to Cineform, but I did end up using Cineform as an intermediate codec for a couple of HDV clips I processed. As a high-quality lossy intermediate codec, it's pretty darn good, (and as a 10-bit colorspace format, it's great for upsampling your 8-bit footage for color correction). It's just not a great editing codec; at least with my current software/system.
  • I need more memory.
  • I need more/bigger hard drives. There is no such thing as "too much free hard drive space", but if I'm ever going to work on a documentary again, I will need a lot more hard drive space.
  • You really can edit and finish an H.264 project in Premiere Pro CS3 with the help of After Effects.
  • I really need a faster processor - render times for the After Effects conform were around 4 hours for 12 minutes of footage. Granted, this is with Magic Bullet Looks applied to every clip and over 60% of the short going through Neat Video processing, but still...
So in all, making this short was a great experience. For my next project, I hope to get a system upgrade so I can run CS6, and be able to edit H.264 footage natively (Along with Red, Alexa, F3, Canon's higher-end codec, etc).

Which deinterlacing algorithm is the best? Part 1 - HD interlaced footage

Before I began, I'd like to give a special thanks to Aleksander Kozak for his help in testing and providing the HD footage used in this ...