Pluraleyes w_no TC_and bad scratch audio on cam.
I'm walking into a project that's almost finished, and they've been shooting on 5d's and 1d's, without TC, sometimes without external camera booms, and without slates. All audio is recorded separately, on 552's (without TC), and sometimes on 788's (with TC). These are usually two camera shoots without any mechanism except pluraleyes for synching recorded audio to what is sometimes very bad camera audio.
I made my argument for strapping a smart slate to my chest and asking shooters to just shoot me whenever they roll, but they said "naw, we just pluraleyes it, and it works fine"
On a later conversation, I found out they hadn't yet pluraleyes'd anything.
Granted, much of what they shoot will be B-roll for interviews that WILL use slates and TC, but I still think they're in for a very rough post.
They're nearing the midpoint of shooting a one-hour to two hour-long doc.
I like them a lot, they're good clients, they pay well, and their finished product has always been very good.
This is, of course, their first time working with cannon dslr's, and they have been told by many that their post workflow will be very tricky, based on how they're shooting and recording audio.
I've used pluraleyes a tiny bit, and I've found it particularly tricky when more than one camera is used, and when recorded audio has a bad track (like a bad lav on one track and a great boom on the other).
They have a lot of footage. They're meeting next week to discuss the post production workflow (I know, I know - would have been good to discuss that before they started shooting).
Since they're not getting it right on the front end (with good scratch aud to cam, nor TC), I've told them they'll have to do extensive pre-organization in post, so they aren't paying an editor to untangle the mess. They asked my opinion of the best workflow now, so I told them I'd write up a page for their meeting.
I've drafted a page of what I'd recommend, and I'll paste it here. People talk about Pluraleyes as being much easier than is my experience of that very impressive and powerful product. It is a great thing, but based on reviews, I expected I could magically dump everything in the same timeline and click a button, which did not work at all for me.
They have interns. But the project has a limited budget, which is why I'm recommending transcripting and labeling files before synching. Please let me know if you think my approach is too conservative, or if you think my workflow here makes sense based on the fact that they're very committed to the way they're currently shooting it. I've also attached - w/format, in case that's easier:
[b]Using Pluraleyes without great scratch audio nor timecode recorded to camera.[/b]
I find that Pluraleyes works best in smaller increments, when combining individual takes of audio and video, or when a shorter series of takes (audio and video) can be paired before pluraleyes’ing.
I also think that Pluraleyes can not take an entire day’s work of video (1 or more cameras), and an entire day of audio (started and stopped at different times than the camera), and magically match them up.
On larger, complex projects, Audio and Video takes often need to be recognized as pairs first, before attempting pluraleyes.
This workflow is further complicated when the camera audio is substandard (as on all cannon cams) and an external boom mic is not attached (nor external signal sent) to the cam for better scratch audio.
Another way this workflow can be complicated is when a lav fails or records a lot of distortion - pluraleyes will struggle to match distorted audio with the cam’s audio. In normal circumstances, if a lav fails, the boom mic can be selected and used as the alternative - if the audio isn’t clean (distortion on lav trak), then that audio should be redacted (but not necessarily discarded) before “pluraleyesing”.
For Pluraleyes to work effectively, pre-organization of audio and video is required.
Transcription is often cheaper than editing time; try not to employ editors for pre-organization.
a. In this case, audio will seldom be used without video
b. Audio usually runs longer than video (especially when there is no slate, and audio does not know when cam is running or not).
c. Since takes will be chosen based on Video selections, you can weed out a lot of audio in advance by knowing what will never be used (i.e. no need to mark/script audio that isn’t in the selected video).
I. Create a clean, untouched backup of all media.
II. Transcribe VIDEO FIRST so you can find approximate synch points to audio.
You can also choose all video selects FIRST, to further simplify this process (only transcribe takes you'll use).
1. Create a transcription (or just a script, since there is no timecode) of all video takes first.
2. Create Take Numbers for each take: 001, 002,... 767, 768... number takes from 001- infinity, in chronological order. If more than one cam is used, label them 001A, 001B... 002A, 003A, 003B... etc.
3. Try to label A and B cameras with matching take/numbers. If B cam was not rolling during A cam’s 2nd take, skip it, and label it with the take that matches A cam (take 003, in the above example).
4. Locate and label audio takes that match (word for word) video transcription. Boom track often works best for reviewing A cam video since boom often follows A cam, and it’s usually cleanest; label that audio with appropriate take numbers.
5. Label B cam take numbers that match A cam takes.
Bring matching audio and video takes into NLE and pluraleyes.
Does this sound like the most reasonable/cost effective way to proceed on this very large project?