Do you really need to transcode h.264 DSLR files?

H.264 is highly compressed and not designed for editing. Because of the high degree of compression, it takes a lot of processor power to play back - especially if you have multiple tracks and effects. If it plays back on your system though, it's fine.

As an ex pro stills photographer, Like you, I'm puzzled by the colour aspect. Maybe someone will chip in with an explanation.

Andy

I agree with Andy in that if it works well for you, then use it.

Since being a teenager (a few years ago) I have usually sought the "best" and most recommended way of doing things in every aspect of life.

Frequently the best way has involved more complex and time-consuming procedures which for my requirements have not produced any noticeable improvement . . . occasionally the reverse.

No - it's not quite like that. When you transcode to PR 422 there's inevitably a tiny loss of quality (PR is very lightly compressed) but the loss is so small, it's not visible to the naked eye (in ProRes 422, let alone ProRes 422 HQ).

I've already been through the thought processes you are following now and as an ex-stills man, I know exactly what you mean but I've come to love ProRes.

I believe what I'm told about colour correction with ProRes, but I have yet to be fully convinced, so if you can play your h.264 without dropping frames, I'd say try it.

These days, I'm using Proxy for just about everything, and then switching to "Original or ProRes 422" on export. I'm very happy with the quality I get.

Again, speaking as a stills man, what bugs me about "ProSumer" video cameras is that the lenses typically have "iris" settings of f1.9 to f22 but if you do simple lens tests, they are totally unusable at anything over f4.5. Why are the manufacturers allowed to get away with this?

Andy

Working with higher bit depths would definitely help with any compositing/colour correction or other manipulations of the original video. This is due to the rounding errors in the math. With low bit depths, you throw away a lot of precision anytime you do any calculations with the data. If you don't alter the video footage, then no, you won't see any difference between an 8-bit video file and a 10-bit that was converted from 8-bit.

In the audio world, the same thing applies. Let's say you record both a 16 bit track and 24 bit track into Logic Pro. They will both be processed (equalizers, compressors, reverbs, mixing, etc) at 32 or 64 bit. While the original sound of the 24 bit track will start with more precision than the 16 bit track, whatever further work is done to them afterwards will be of equal precision, as they're both processed at the same bit depth. The reverbs will be of equal quality, as an example.

So, for video, if you were keying, converting an 8-bit video to 10-bit won't make your key any better, because you're relying on the precision of the data you already have. However, if you were compositing smoke over a background shot, the composite math would be more precise (detailed?) at 10-bit than 8-bit, as you're actually generating new data. It would also be better for colour correction, as even though you're starting off with less data, the computer can create interpolated data to work with, let's say for mattes and selections. That wasn't super clear. Sorry.

So, the question here is, how does FCPX process different bit depth video files? If you have your project setting as regular ProRes, but are using original 8-bit camera files without optimizing, will FCPX do the calculations exactly the same as if you WERE editing from ProRes? My thoughts would be yes, it processes everything (effects, alphas, colour correction) in 10-bit from the get-go. In which case, it doesn't matter, other than whether your processor can handle it, whether you optimize or not. It's still better to initially RECORD in 10-bit, but once the 8-bit damage is done, might as well stay with 8-bit files until final output, while letting FCPX do it's processing in 10-bit. Hopefully FCPX doesn't stay 8-bit through any of the processing, converting to 10-bit later on.

Anyone have the definitive answer?