Here's the problem:
Putting SD footage in a 4K timeline. This forces a double conversion, which severely degrades image quality. SD footage gets upscaled to 4K during rendering, which means that the SD image gains no more information (but just enlarged instead), and then downconverted back to SD during encoding. That second downsizing causes a severe hit in image quality! In other words, that SD footage has only 480 horizontal lines of information (or a vertical resolution of 480 lines) even when blown up to 4K size - and then, downconverting that causes a resolution loss by the same factor as downconverting native 4K footage to SD. Assuming that the 4K source has a (hypothetical) resolution of 4096x2304, the SD footage so processed now has an effective vertical resolution of only 100 lines! The SD clip in question has already lost 96 percent of the image's "original quality" (or 80 percent of the horizontal and the vertical resolution) right there.
This project is better done by creating a new SD timeline (such as a preset used for DV NTSC Widescreen) and then putting both the 4K and 480p files into that SD timeline and then using the "Scale to frame size" option (right-click on the video portion of the 4K clip on the SD timeline, then select "Scale to frame size") on the 4K clips on that SD timeline.
In other words, if you're working with different clips with mixed resolutions, it is best to use a timeline sequence that exactly matches the resolution of the final output. This will prevent double-scaling-conversions (upsizing during rendering and downsizing during encoding).
By the way, what are the exact specs of your computer system (CPU, RAM, GPU, hard drive(s), etc.)? (I assume that you already have a powerful enough CPU to handle 4K footage, especially since it requires a robust CPU just to handle it satisfactorily.) Underspecced or underpowered PC hardware will result in excessively long times for playback, rendering and exports. You might not have enough fast disks installed inside your system, or you might have a weakling GPU or a GPU that cannot use the Mercury Playback Engine's GPU acceleration at all.
If you can't get the original clips before they were sent to youtube, you might try some scaling software to preprocess the clips before editing.
Also keep in mind what is your targeted output, DVD, BluRay, HD or SD broadcast, or web media. If you are not exporting to HD, it might be wise to edit in a SD sequence and deal with the long export time.
Also you could edit a RED sequence natively, leaving place cards for sd footage, export to a high quality SD codec, then insert the SD quality youtube clips in a new SD sequence. Sounds like a bother.
pulled from Youtube as MP4 quicktime videos.
I think that if this is not the problem, it is a contributor. YouTube Transcodes submissions, and those are usually pretty highly compressed. When you add the compression in the MP4 (assume H.264), things can get pretty ugly.
I agree with the workflows stated by wfmc staffer.
Thanks so much for the detailed response! I really appreciate it.
A few questions going off what you've posted.
- So does in the monitor window of Premiere the video clips look okay because it's showing at a smaller resolution, as the window size is smaller correct? So it's not a good estimate of how it will look on export? We'd hooked up the computer to a 1080p monitor and played full screen and the 480p clips looked okay, even scaled up to 1080p. But I assume Premiere isn't displaying the full 4K quality (unless of course you have a monitor that can display that resolution). The 480p clips look okay in the Premiere monitor window and while editing, but that gives a false sense of the export quality correct? Because as you said the 480p clips being blown up to 4K are extremely degraded.
- What I had done was create DV NTSC Widescreen sequence (switching interlaced to progressive in the sequence settings, and from 29.97 to 24fps), and nested the 4K sequence in this sequence, and scaled the 4K sequence down to 21.5% to fill the 720x480 widescreen sequence. But would you recommend just copying and pasting all the clips into this sequence and selecting the "Scale to frame size" instead of nesting the 4K sequence? Would this potentially yield faster export times? I had assumed nesting the sequence would be the best option, but perhaps copying and pasting the clips themselves into the DV NTSC Widescreen sequence would be a better way to go about doing things.
I'm thinking that perhaps the hard drives are to blame as well, as even though they are internal the footage is accessed from all the drives and we were writing to one of the drives - maybe writing to an external drive connected through eSATA would be faster.
Thanks so much!