So I finally got around to doing some compression tests. My source was a 1280×720 23.98fps 10bit 4:2:2 “uncompressed” quicktime. That is a typical file that comes across my desk. I wanted to see what compression I should set my Final Cut Pro timeline to. At first, I opened the quicktime in Quicktime Player and exported to a variety of codecs. But when I opened all of the files, I noticed that they were all darker than they should have been. That indicated to me that Quicktime Player had done some sort of colorspace conversion. So I redid the same test from inside Final Cut Pro. Sequence Settings were set to the same as the source video.
For the test, I used these codecs:
- None
- Animation (100%)
- PNG
- Apple Intermediate Codec
- MJPEG-a
- MJPEG-b
- HDV
- DVCPRO HD
- Prores
- Prores (hq)
- 4:2:2 8-bit
- 4:2:2 10-bit
Results
So I noticed that None, Animation, PNG, Apple Intermediate Codec, and MJPEG went through a colorspace conversion. So I won’t be using them again.
HDV and DVCPRO HD were slightly fuzzier than the original, but also smaller than the remaining formats. DVCPRO HD also underwent a slight tonal change in image quality. DVCPRO HD also has poor cross-platform support, and both of these formats have restrictions on their frame size.
Prores lived up to its name, and was as fast to compress as the 4:2:2 “uncompressed” formats, while also yielding files over 75% smaller, with little perceptual quality loss. They also had the benefit of not causing FCP to re-render the timeline.
<needs more pics>
Conclusion
I’ll have to re-run this test with a simple filter applied to the image. I’ll probably add 8 pixels of black to the bottom or something. I think that 4:2:2 10-bit won the fps race because Final Cut Pro just copied the data instead of having to recompress as it did with all the other formats. I also haven’t figured out why Prores HQ was faster than Prores.