Audio Workflow on Pizza Testimonials
 

I got some great feedback from my East coast pizza testimonial videos and a lot of people requested more info on how I handled audio. I put together a quick tutorial on my workflow.
http://www.joncarrpro.com/blog/2009/...-workflow.html

Here is a link to the final videos:
http://www.joncarrpro.com/blog/2009/...timonials.html

Good tutorial - one note though is that the 99.9% speed adjustment is no longer necessary to maintain sync if you create an easy setup with the correct timebase, as described on the pluraleyes blog (not specific to pluraleyes):

25 Hour Day: DSLR Dual-System Audio: The 99.9% Solution

Related to that - based on my own testing I don't believe the part about the 5D really shooting at 29.97 is true - changing the file headers just results in the footage going out of sync with it's own audio. The problem is entirely with how FCP treats audio timing on import and the easy setup fix eliminates the problem.

Excellent writeup John. I didn't realize NeoScene gave you the Cineform codec. Thought you had to buy the $800 package or something to get it. Maybe that was for the 4K one.. lot of options on their site. I'll have to check that out as I would rather work in the CineForm format as well.

Do you edit any background noise out of the sound later on? I thought I heard some background hum noise or something.. but in the video spot it didn't sound like that. You don't show that in this tutorial so wasn't sure if you take that clip just as you did it there, or if you clean it up a bit more.

I saw that blog posting and tried it but I still had drift issues on this project. I will probably do some more digging and update the video.

I did clean it up a bit after.

Just so I'm clear on this, does cineform leave the file as 30.00? If so, than wouldn't fcp conform the video to 29.97 in the timeline. (given that you are in a 29.97 timebase sequence.) I ask because it seems like if the video is converted to 29.97 BEFORE you drag it into the sequence, than you wouldn't have issues with the audio sync at all, because both 5d video and separate audio would be running at the 29.97 rate.

Sorry if I'm over-complicating things!

As I understand it, the newest version of Cineform DOES convert the files to proper 29.97.
There have been a few threads here previously and there are more in the Cineform forum here on DVInfo.

Steve, the 5D MkII shoots 30p and records in real time. The latest version of NeoScene changes the header to 29.97, and it resamples the audio to slow it by 0.1%. Drop the video on a 29.97 timeline and it will play both audio and video slowed down by 0.1%.

This is all good until you try to match the 5D with a 29.97 camera, which plays in real time. The 5D will be slower that the camcorder by 0.1% in both audio and video.

Did you try it on a completely new project? I continued having problems too until I started completely fresh in a new project and then it worked perfectly. It seems like once FCP has a timebase in it's head for a project it's not possible to really change it.

Here's what I found - I'm using ProRes & H4n audio, but I believe it should be applicable because this is a general FCP issue:

Outside of FCP the 5D video and separate audio have the correct equivalent durations (in QT's movie inspector).

Once imported into FCP, in the project window the audio has a shorter reported duration than was reported by QT.

If you place the audio into a 30p sequence it retains this shorter duration. 5D footage added to the sequence has the correct duration and is therefore longer than the audio and out of sync. Slowing the audio to 99.9% stretches it back to the correct duration and returns sync.

If you place the audio into a 29.97 sequence the audio duration returns to that reported by QT. If you add your 5D footage to this sequence it is slowed down to conform to 29.97 and ends up longer than the audio and out of sync. It doesn't matter if you let FCP conform to 29.97 on the timeline or do it first in some outside application - the end result is that the video will be longer than the separate audio tracks.

After you follow the directions in the blog post I linked above and create an easy setup with a true 30 fps timebase, import your audio and it will report the same duration as QT reports externally. Drop it and the camera footage into a 30 fps or 29.97 timeline and both the video and audio retain their correct durations and sync.

Since this seems to be a problem with the way FCP imports audio, once it's imported incorrectly it's too late to change things. That's why you have to start a new project using the corrected easy setup for this to work. I haven't tried copy/pasting from a bad project to a good one to see if that fixes things yet.