I do alot of back-and-forth between AE and FCP; bring DV footage into AE, adding text and other graphics and sending back to FCP for further editing with audio (and then final compression with Sorenson Squeeze). Recently I've been doing my intermediate render with Photo-JPEG compression as it seems like my file sizes are much smaller than with the animation codec and quality doesn't suffer. I havn't found too much online regarding Photo-JPEG although I've heard that sometimes playback smoothness may not be great.
I'd appreciate any advice or comments on this codec, or compression workflows in general.
I'd like to know the pro standard for going back and forth with AE/FCP, or is it kind of a 'whatever works for your situation' deal?
We have somewhat abandoned Photo-JPEG for an entirely DVCPro50 workflow. We think the file sizes are manageable and the codec holds up to abuse (layers etc). If we are working in uncompressed 8 or 10 bit we stick to that codec in between.
Disk space has become so cheap it does not make sense to do as much transcoding as we might have done in the past. We also prefer the compression of DCT type codecs rather than the destructive Jpeg codecs. Although PhotoJPEG is a respectable codec, it can create color shifting when brought between apps.
PhotoJPEG at 99% or lower can introduce quite a bit of 4:2:2 chroma filtering. At 100% quality there's debate whether it's truly 4:4:4 but the quality is pretty pretty good, but not as pixel perfect as Animation. You might want to try PNG as an alternative. With images that contain a lot of solid colors, it can save a lot of file size space.
To see these examples, click the Codec Resource link in my signature.
I have a rule I use when rendering either RGB 4:4:4 Lossless (like Animation codec) or direct to 4:2:2 Uncompressed (like 8-bit Uncompressed)...
8/10-bit uncompressed - if the rendered clip will go to the NLE timeline (in my case, FCP) *and* there wont be any layers above that new clip in the timeline (that will further be rendered again).
RGB Lossless - if the rendered clip will go to the NLE timeline and I'll add more video track layers to it (like text). This ensures the original clip stays pixel perfect and basically performs a first generation encode when it and the new layers in the NLE are rendered.