[Ben Withington]

Thanks for the feedback, I’ll check out the renderer settings.

However, I think I’ve found the direct problem. The tutorial I used to learn the rotobrush failed to cover the “freeze” function. So what I’ve done is turn off all the other layers, effects, and frame blending except for the rotobrush effect and layer. I hit freeze and it started propagating a lot faster with lower numbers (I’m guessing the frame blending was adding that extra work). So I was applying color correction and motion blur to an unfrozen roto mask that was frame blending itself live for every frame.

[Ben Withington]

If I did a RAM preview of the whole composition in AE, just the problem layer, or are you saying something else?

[Ben Withington]

Thanks Michael, that’s very helpful. Gotham Sound also verified for me that they sell the unattenuated version unless otherwise noted. At this point if I need particularly specialized microphone setups, I’ll just rent.

I’m putting together my starter package and am gravitating towards getting Gotham’s combo with the G3 packs.

[Ben Withington]

Wow, thanks Dave! I had no idea 422 was that durable, that will significantly ease up my workflow. My noobie speculation faculties were generating paranoia about my image falling to pieces if ever projected on a big screen. Obviously it will never be RAW or higher than 1080p, but at least I know now that I was worrying about nothing.

[Ben Withington]

Found this post from Dave LaRonde at the Adobe forums:

“If the footage you imported into AE is any kind of the following — footage in an HDV acquisition codec, MPEG1, MPEG2, AVCHD, mp4, mts, m2t, H.261 or H.264 — you need to convert it to a different codec.

These kinds of footage use temporal, or interframe compression. They have keyframes at regular intervals, containing complete frame information. However, the frames in between do NOT have complete information. Interframe codecs toss out duplicated information.

In order to maintain peak rendering efficiency, AE needs complete information for each and every frame. But because these kinds of footage contain only partial information, AE freaks out, resulting in a wide variety of problems.

I’m a Mac guy, so I like to convert to Quicktime movies in the Animation or PNG codecs; both are lossless. I’ll use Apple’s Compressor, Adobe Media Encoder or Quicktime Pro to do it.”

I am using H.264, and that turned out to be the problem. I converted to 4444 and it immediately worked.

However, I’m a bit bummed because this adds another generation of compression to my workflow for this particular project (I edited it in native H.264 in FCPX, so that has created some difficulties for myself). I also want to use a 3D Lut, which will not work in FCPX.

If I can convince myself that H22 (HQ) is ok for festival screenings, I could do everything but the grading in FCPX and that’d streamline things.

[Ben Withington]

It might be worth mentioning that the effect does work properly in FCPX. I have a Nvidia 9600 GT and in the preferences for Neat Video it says, “No computation -capable GPU devices found”. This is evidently because of the plugin not being able to load the CUDA driver for my GPU. However, from my research it seems that feature is not required for the plugin to function as it should; it’s more of a performance thing.

I also tried re-installing the plugin which didn’t change anything.

As far as I can remember, all the settings in AE are default for a CS6 installation. I’ve only been changing the project to 16 bpc, but tried it in 8 bpc too just to try it.

[Ben Withington]

I think I forgot to mention that I’d like the footage to hold up as best it can on a big screen in HD projection. If this was online or disk distribution only, I’d definitely go with ProRes. I admit this is out of speculation on what little I know and it could be way off base, but my reasoning is as follows.

422 (HQ) guarantees no compression artifacts to be visible on the first generation; evidently there is quality loss even though it’s imperceivable. However, since my source is highly compressed H.264, I would have three generations of compression in my workflow: the H.264 in camera, the transcoded format for color correcting, and the final export. Even though artifacts can’t really be seen in either iteration of 422, I wondered if that holds with the kind of scaling that happens with projecting for a larger audience than a 1080p monitor.

So, I thought it might be a good idea to use the animation codec. This way I know there is no quality degradation regardless of whether I can see it or not and the final export will be as close in quality to the original H.264 file as possible and I have more wiggle room in the compression of the final export depending on a venue’s specifications.

The only problem is that I couldn’t edit animation files cause they’re too darn big. I just needed to make sure the proxy media would show me the correct responses to my color correction so I wouldn’t get surprises when reconnecting. Thanks for testing the process out by the way, I don’t know why I couldn’t figure out how to check the color space of the proxy footage.

So that’s my theory anyhow. Definitely let me know if I’m making more work for myself than I need. I certainly wouldn’t miss the extra step of using After Effects, but absolutely making sure the quality stayed as close to first generation as possible seemed important to me for this kind of project.

[Ben Withington]

That FCPX is 64 bit is referring to the fact that it is able to take advantage of 64 bit processors. When I say 16 bit, I mean 16 bpc or 16 bits per channel, referring to the color channels of the footage. I think the idea is that upconverting doesn’t changing anything within the image, no, but it does give you head room to preserve pixel quality and have cleaner results form the information that is added to the image by color correction and grading.

I don’t mind compression artifacts when I’m correcting and grading, I just want to make sure that what I’m seeing in terms of color and contrast ratios will be identitical when I reconnect. You know, like pushing the image to where it posterizes the proxy media but not the original media.