I transcoded my Canon 5D footage to ProRes 422.

Edit is locked and ready for color grading.

Should I colour grade the ProRes files in my timeline or should I replace the ProRes files in the timeline with the original Canon 5D files first and then color grade?

Grade the ProRes files. In short, you're buying yourself more color latitude and getting shorter render times in the process.

interesting article on colour correction..

Color Correction: Put Your Best Foot Forward | Hurlbut Visuals (http://hurlbutvisuals.com/blog/2010/03/30/color-correction-put-your-best-foot-forward/)

I agree that to expand range is better.. but.. one question.. If so.. why ProRes 422 and not Prores 444 ?

Grade the ProRes files. In short, you're buying yourself more color latitude and getting shorter render times in the process.

To clarify; you cannot add color depth, latitude or dynamic range to the original footage by transcoding. What you gain is more control with less artifacts and better compression. Also important is to grade using a high bit depth, i.e. 10 bits and not 8 bits. Adding a little den-noise can help because you will create new color values for some pixels.

ProRes is a near lossless codec, so you wont spot any degradation using it for grading or FX, H264 starts to fall apart pretty quickly.

As for why 422 and not 444, do a simple test rendering the same footage to both. If look at the footage you probably wont notice a difference, but if you look at the file size you certainly will, that's why there's a choice.

Obviously.. filesize is bigger.... but.. if the idea is to have more colour space etc...

The key to a good color grade is to put your NLE in a high bit depth mode.

This weekend we did a shoot in which the opening image is a closeup of a face that is out of focus. We shot it at 1200 ISO with the latest firmware are 24p. The Picture Style was Neutral with Contrast and Sharpness and minimum and saturation somewhat reduced. We transcoded with the latest version of NeoScene and edited/graded in Vegas Pro 9.0.

As I recall, our plug ins were:
* Luma color curve for fine tuning the exposure
* Glow (for a very slight diffusion effect)
* Moderate desaturation.
* RGB color curves for color grading, pushing the shadows toward teal and the highlights towards salmon.

Overall, we were going for a horror film look.

With the Vegas project set for 8-bit processing, we get the dreaded contour lines. That said, the latest firmware really rocks. The contours and smooth and not blocky in the least. But it still looks more like a topographical map than a blurred face.

Then we set the project to 32-bit float with studio (16-235 equivalent) levels. We set the preview to Best, Full and viewed at full screen.

The resulting image is gorgeous. You couldn't find contour lines with a magnifying glass. There is a slight grain and texture, but no blockiness.

My conclusion is that many people who complain about 8-bit video are actually having problems with 8-bit processing of video.

Back to the original question, I did this grade with transcoded Cineform clips. And I'm confident that I would get the same, excellent quality had I used the original MOV files. I believe that the keys are to add some slight pixel-to-pixel processing, like diffusion, softening, or noise reduction, and to use 32-bit float or equivalent processing.

Given the improved speed for editing, I recommend using transcoded video, given a high-quality, 10-bit or better transcoder, like Cineform.

BTW, with Vegas, we do all of the editing and correction with 8-bit processing, except when pixel peeping and for the final final render. Performance really suffers on my old, tired PC in 32-bit mode.