Lossless MJPEG exports?

We've just finished shooting a feature in MJPEG at the resolution of 4096 x 2160 with a Canon 1DC and are about to start cutting with FCPX on a Retina Macbook Pro with 16GB of RAM - so far viewing the rushes and making preliminary cuts has posed no problems and we're particularly pleased there is no lag or slowing down of the footage which simply plays like it was any other smaller format…

I have a question about MJPEG and FCPX and my own cutting methodology. When cutting, I like to copy quite a bit, so to speak. For instance, I will focus on a particular scene, cut it together and render it. Then will take the rendered file and design its sound, lay it all down, then render that. I will do this, individually for each scene in the film. Then I will dump the fifty or so scenes into a fresh timeline and render that also. Subsequently, I will lay in the score and re-render. So, as you can see, there are many generations of copying / rendering / exporting.

In the old days, I would have been going back to the neg or in analog video using a timecode and go back to the masters… but I prefer not to go through a similar process here. Therefore, what I need to know is whether the duplication I have outlined above will cause any degradation to my image whatsoever. To be totally clear, it's all MJPEG files at 4096 x 2160 and every single time I export something I am religiously doing it with all the best, highest, slowest settings and outputting at 4444 and so on - with so much reverence that it's almost as if I am handling a negative.

But is it possible I will lose anything from the image, even the slightest bit of quality, doing what I have described? If so, I will have to find another way. I would really appreciate any advice (or reassurance?) you experienced FCPX folks can throw my way, thank you for it deeply in advance and ask you to please remember this is not a request for help with cutting methodologies and so on (I have that part figured out over the years) but rather a straightforward technical query about whether repeatedly exporting MJPEG will cause the image to slowly, even slightly degrade, thus making my methodology a faulty one!

Is there, for instance, a particular setting or settings that will ensure the MJPEG remains as MJPEG until such time as I want to create various versions of the master later on?

I need to be darn sure that my master is 100% for many obvious reasons, not least of all that we'll be premiering in 4K cinemas.

Since originally posting I have learnt a few things...

Firstly, I've discovered how robust the Pro Res 4444 is supposed to be. It seems that if I do what is described in my initial post, each time rendering to Pro Res 4444 with all the settings at the slowest and best, it's really unlikely any degradation will be visible to the human eye. From what I can tell, after the first render, I will be in a Pro Res 4444 context and further degradation is not really, effectively, going to occur at least in terms of human visual perception. Some say you can copy the material 10 times without noticing any difference, and Apple themselves seem to imply more, almost as if quality loss just ceases because your using Pro Res 4444. I would like to know if this sounds right to you all, or whether it's marketing/fanboy stuff?

Secondly, I've learnt about an FCPX plug in called, ironically, 'Compressor', which allows for straight file-copying type results, in which the 1s and 0s of each frame are preserved perfectly, and only the actual order of the frames is changed, depending on your edit choices - theoretically this should be possible with MJPEG because the compression is intra frame compression (compression within each image) and not inter frame compression (compression that works between images, often meaning individual frames are 'missing' or guessed etc.) but I would love to know if anyone has used Compressor to create lossless or even just used Compressor at all...

Finally, I've been told rather forcefully that rendering and exporting is always lossless and there is no such things as lossless NLE cutting, particularly when using a 'lossy' format like MJPEG - the person saying this took the position that it doesn't matter in the slightest that MJPEG is intraframe, and claimed it's impossible to simply re-arrange the frames intact, because there will always be recompression...which doesn't tally with my experience and is something else I would like your thoughts on?

Hi Pierre, it seems to me that you are quite cautious with the quality of your footage. And that makes me wonder why you do all this rendering/exporting/re-importing in the first place? I mean, your footage taken with the 1DC comes as QuickTime Motion JPEG. Now that's quite some compression, before you even start to edit. But--and for me that's the important point--it is the original media and will reside in that Original Media folder untouched, unless you export it. And if I would want to keep the best possible quality of my footage, I'd export it only once: At the end of the race. Once you export you bake any loss into your footage and bring it back to your edit with the re-import. So I'd leave everything within FCPX. But that's just my 2 cents.
(By the way, if you really stick to the ProRes4444 format all the time after the initial rendering from JPEG to ProRes444, I also think that there won't be any recognizable loss of image quality. Other opinions appreciated; interesting topic.)

Yes, maybe I am too prejudiced against the idea of conforming at the end. Conforming used to be hassle. Maybe it's easier in this new FCPX context.

Someone else has suggested meanwhile that "if you leave the footage alone in FCPX then any exported file will just be a simple reordering of the frames" and this really engages with the spirit of what I am trying to say - and is something that I may not have made clear enough in my initial post.

In short, I don't really mess with the image when cutting. I generally favour straight cuts and don't go in for effects. Even grading won't be a major process for us, as we favoured Canon Portrait over Canon Log and got the results we wanted in camera. Obviously we'll grade, but only where problems occur.

I am hoping that by rendering to Pro Res 4444 I will 'catch' the quality pretty much at its original level and not degrade it thereafter, which is to say not encounter degradation at all in this process really.

It seems the answer to my original question may be something like: technically, there is loss but in reality there is not!

Pierre - In your OP you mentioned exporting for sound then importing back in... why not export an XML and use x2pro to send to audio post then import your final audio tracks into FCPX and use roles for your exporting. I would also export an h264 for video reference only for audio post. This way the video never leaves FCPX and never gets compressed or re-compressed time after time during your exports for audio then your final. You could also use those same XML's to color grade in DaVinci if you go that route (or stay in FCPX for grading).

ProRes HQ would do you just fine. ProRes was designed to, and in fact does, hold up to no generational loss up to 10 generations. There's no worry about transcoding with it.