While H.264 is more efficient than MPEG-2 for instance, providing decent quality at low bitrates, it is a lossy codec. Any time you re-encode your video, there is a quality loss. If possible, your best bet is to encode to various output formats direct from the Premiere sequence, using the original source material, so you are going from the camera clips direct to destination clips, with only one transcode taking place.
If you must create a "master clip" first that you will later create other files from, then you want that master to be a high-quality format. "Uncompressed" video takes up crazy amounts of space, so you'll want to find a codec that is considered "visually lossless", which will provide a high quality image with manageable size. Note these files are NOT for upload/web, they are too big, these are for keeping a master clip on your hard drive for archiving.
On the Mac, use Apple ProRes. On the PC, you can use the Avid DNxHD or Lagarith codecs, both available at no charge. You might also look at the Cineform codec. There's other options, but these are popular.
Safe Harbor Computers
Hi Jeff - thanks for the feedback.
In this case, if you are going from H.264 using a sequence created from a native clip and then selecting same settings as sequence on the export, does PP encode the video portions that haven't changed? I have a number of titles in a second video track and crossfade transitions between each clip and I realize those would need to be encoded, but what about the rest of video? Just curious how it works here.
While there are some NLEs that do in some cases offer "Smart Rendering" options, with Premiere, any export is always completely re-encoded.
Therefore, you want to keep re-encodes to a minimum to avoid quality loss.
Got it. Before I switched over to PP I was using VideoStudio, a consumer app, and it had a SmartRender option which I was guessing might be the case with PP (and have now found it is not).
One seldom mentioned aspect of "smart rendering," is that it only applies to footage, that has not been changed in any way. If one adds any Effect, the footage will need to be Transcoded. If one adds an overlay, such as a PiP (Picture in Picture), or Title - Transcoding will be required. Even with most Transitions, the area of "overlap" will need Transcoding.
If one just does Butt-Cuts, then the output will not need to be Transcoded.
That comes as a surprise to many.
That is exactly how VideoStudio worked - only rendering what changed. That's why above in my original post I mentioned that my project was farily limited in video stream changes, just some titles and transitions. However, VS does such a horrible job encoding HD that I don't think we should be using it as a comparison :-).
I may put in an ehnacement request, but in the meantime will first play around with a couple of the lossess / near lossless codecs.