[Ivan Myles]

There are several compression parameters that interact in different ways depending on the encoder settings, so it is difficult to give a straight answer. In general, the I-Frames (aka key frames) require more data, and the P- and B-Frames are smaller. In simple terms the I-Frames are broken down and spread over several cycles in the bit stream.

The degree and methods used to compress the frames are determined by the bit rate, profile, level, and other settings (depending on the codec and encoder). In multi-pass encoding the compression engine can optimize the data to a greater degree.

If you want to conduct a visual experiment, encode a short clip using a variety of encoder settings: 1-Pass, 2-Pass, different profiles, levels, and bit rates. Export maybe 8-12 versions, then import the files into Premiere Pro and stack them on top of the source sequence. Apply the difference matte effect to each one and compare how closely the compressed files resemble the source clip. (Here is an example, but it is easier to view frame by frame.)

For a direct answer to the question, “How does the first pass analyze the clip?” there is a wide variety of reference material available on H.264 encoding. It depends how much time you would like to devote to it.

[Ivan Myles]

You might be able to get acceptable results by using After Effects to interpolate new frames. Take a look at Timewarp and Frame Blending with pixel motion.

[Ivan Myles]

Based on your comments it sounds like the source video is not well suited to high compression. The artifacts you described are examples of starved bit rate. YouTube re-encodes uploaded videos to about 0.07 bits per pixel (bpp). By comparison, the “still lousy” 16 Mbps H.264 file you rendered is 0.27 bpp, and the “somewhat better” uncompressed 422 file is 16.00 bpp. One option is to post a high bit rate H.264 on a different web site, but ultimately the animation should be simplified to look better at low bit rate. Please refer to the following post for some suggestions on simplifying a video for YouTube: Re: Terrible YouTube compression

[Ivan Myles]

[Gary Verby] “Depending on how precious your cuts are I would get into the habit of exporting mastered (uncompressed) versions to combat this issue down the line.”

If storage space is not available for uncompressed files, consider using a lightly compressed production codec like ProRes, AVC-Intra, or DNxHD/MXF.

[Ivan Myles]

Here are a couple of points to add to Kylee’s sound advice:

– If you plan to export from Premiere Pro, consider pre-processing (or, more correctly, parallel processing) the source files in AE rather than going back and forth between the applications.

– Music rights can become tricky depending on how you want to share the finished video. Royalty free agreements typically provide synchronization rights that allow music to be incorporated into a video. Many licenses also grant limited distribution rights. However, this does not necessarily mean the video can be played for other people. Depending on your situation, it might make sense to acquire public performing rights so that you can show the finished video freely.

[Ivan Myles]

Just getting caught up on this thread and I think it wound up in a good place. Ultimately, the best way to understand how different variables affect the end result is to 1) develop a stronger understanding of the compression process, 2) become familar with encoding guidelines/practices, and 3) perform systematic tests. Encoding is well-suited to testing because it is easy to vary one parameter while holding others constant. Over time, compressionists learn how to improve not only the encoding settings, but also upstream activities like editing, compositing, colorization, camera movements, et cetera to ensure smooth playback.

[Ivan Myles]

If you do not select frame blending Premiere Pro will only use two out of every five frames and drop the other three. A series of frames like ABCDEFGHIJKLMNO at 60fps might appear as ACFHKM at 24fps (or possibly ADFIKN or BDGILN).

[Ivan Myles]

Note that DNxHD in a MOV wrapper might experience color/luma shifts and clipping; compare the transcoded output to the source files to check for potential issues (preferably in Premiere Pro with the YC scope and RGB Parade.)

[Ivan Myles]

Also, if you have access to After Effects try using Motion Stabilization instead. I can speak from personal experience that manual frame-by-frame stabilization is extremely time consuming.

[Ivan Myles]

[Richard Boswell] \”I\’ve created a test output of just a single clip using different settings… https://www.filedropper.com/outputtests\“

Sorry, that link does not show the test file.

Are there a lot of quick cuts in your edited file? You can always try encoding at higher bitrates (20-40 Mbps) with a maximum bitrate at least 1.5x the average setting. It might not impact the re-encoded YouTube file, but it will help with troubleshooting.