Post-Audio Workflow for Documentary

[Stephen Brock]

Hey everyone, I’m working as an AE for a feature-length doc and I am wanting to gather some direction before we get too far down the rabbit hole in our post audio workflow.

Like most docs, we are shooting on a variety of formats but at least 85% of our production workflow is C300 footage and 5-channel mono from a Nomad field mixer/recorder. Other times we have 5D’s, GoPro, iPhones, and even some RED from scripted segments.

With exception to the RED footage, we are transcoding all footage to ProRes 422 HQ. But before we begin that process (475+ hours of footage), I wanted to find out the best workflow for audio.

We have a variety of situation/scenarios, such as Broll, formal interviews, variety(reality-style), events, etc. During formal interviews, a mix down was sent to cam A direct input while cam B had a scratch audio mic. During variety and broll the cameras were either scratch-only or no-audio, while the nomad was relied on separately. Timecode is helpful but unreliable since the crew was often in stressful situations and had no opportunity to jam.

All this being said, whats the best practice since we plan on handing off a picture-lock to a post-audio house for mixing?

Here’s my current guess:
1) Import all assets into premiere
2) Set up sequences as 7-channel mono for cam-audio and nomad combined
3) Sync audio to correlating footage (combined techniques of plural eyes, PP multi-cam, timecode, and manual claps)
4) Rearrange channels to ensure consistent assignments (boom on ch3, lavA on ch4, lavB on ch5, etc)
5) Export to ProRes 422 HQ. Keep original audio settings—48khz, 24-bit, 7-channel mono.

This workflow is likely going to be my life for the next couple of months…lol
Thoughts?

 

[Doug Beatty]

The simple solution is to ask your post audio house how they would like you to do it. My best guess is that they will not want re-rendered QT audio and want the original 48/24 Nomad poly files.

A cam audio is a throw away any time more than one source is summed to a mono track (phasy) but B-cam scratch mic may be useful for shots in the field.

Pluraleyes is going to be your best bet since you are using cameras without timecode, splitting your bins by date and not throwing everything at it at once.

 

[Stephen Brock]

Yes of course I would normally just ask the post audio house how they want me to do it, but we are still shopping around for the right company. In the meantime, our lead editor is wanting to begin cutting.

So until then, I have to make do with the workflow most likely to be accepted...

Now if post audio house wants to use the original nomad files, how will they easily relink without timecode references? Should I just make sure all my ProRes QT's adopt the nomad timecode regardless of it's consistency?

 

[Doug Beatty]

The editor may be painting your team in a corner by trying to start right away without locking down your workflow. A few days now could save you a few weeks down the line. Everybody has their own way of working so it's best to be on the same page from the get go.

Juniper Post has a handy document on how to prep your project for OMF/AAF export that could give you some clues to best practices. David is a really nice guy, so it may be worth a call over for advice.

All I'll say is that rendering and re-importing can destroy metadata and metadata is your friend on projects of this scope, so be careful.

 

[D Fuller]

Yes of course I would normally just ask the post audio house how they want me to do it, but we are still shopping around for the right company. In the meantime, our lead editor is wanting to begin cutting.

So until then, I have to make do with the workflow most likely to be accepted...

Now if post audio house wants to use the original nomad files, how will they easily relink without timecode references? Should I just make sure all my ProRes QT's adopt the nomad timecode regardless of it's consistency?

If you are indeed in a hurry, and want some solid advice, I'd suggest you PM Mark Weilage and ask for his thoughts.

 

[Marc Wielage]

My observation is that you should make absolutely sure you keep the Sound Roll Numbers in the edit 100% of the time, and make sure all the metadata is accurate (track names, scene & take numbers, etc.). This will help make sense of the OMF or XML later on, when the sound is prepped.

I would not wait to talk to a post sound house, but instead get on the phone with them right now to make sure your workflow will be compatible with their particular setup (most likely Pro Tools 10 or 11). Make sure you provide the post sound house with sound reports on each sound roll [folder] for each day, so they can locate any missing takes, alternate takes, B-roll, or wild sound.

In a perfect world, your on-set sound mixer should have created some kind of reasonable mono mix to use during the edit, so your editor can disable all the other tracks and just worry about choosing the right takes and telling the story. In a pinch, if somebody's mic drops out, the editor can always grab that one line and bring it up if necessary, but the hope is that they'll rarely need to touch the isos during this stage.

I agree with Doug above that you should not export the audio in any way at this point. The best workflow is to always use the original files, and just allow Pro Tools to reconform the edit and match the picture from the WAV files. It's easy and not time-consuming to do, provided the editor has dotted all the I's and crossed all the T's. I would try a sample 2 or 3-minute scene to make sure it's going to work; Premiere is not my favorite editing program. Most shows that I see worked on are done in FCP or Avid. Premiere "should" in theory work, but it's not always ideal for longform projects.

Timecode is helpful but unreliable since the crew was often in stressful situations and had no opportunity to jam.

That's sad, because often five minutes on location can save hours of work in post, particularly in multiple-camera situations. I think the best way to do this is have the sound department strap on lockit boxes (or Denecke SB-T's) onto the cameras, and those are good for 5-6 hours without losing a frame. It makes it much, much easier for the editor when all the cameras and sound precisely match in timecode.

 

[Frank Weeks]

because often five minutes on location can save hours of work in post,

Amen