Comment by socalgal2

The issue is encoding is an art, especially as it's lossy. You choose how much data to throw away (kind of like when you pick a quality in JPEG). Further, for video, you generally try to encode the differences between 2 frames. Again, because it's a lossy difference, it's up to the creator of the encoder to decide how to compute the difference. different algorithms come up with a different answers. There result still fits the spec.

Let's just say we were encoding a list of numbers. So we get a keyframe (an exact number) and then all frames after that until the next keyframe are just deltas. How much to add to that keyframe number

    keyframe = 123
    nextFrame += 2   // result = 125
    nextFrame += 3   // result = 128
    nextFrame -= 1   // result = 127

etc... A different encoder might have different deltas. When it comes to video, those difference are likely relatively subtle, tho some definitely look better than others.

The "spec" or "codec" only defines that each frame is encoded as a delta. it doesn't say what those detlas are or how they are computed, only how they are applied.

This is also why most video encoding software has quality settings and those settings often includely the fact higher quality is slower. Some of those settings are about bitrate or bitdepth or other things but others are about how much time is spent looking for the perfect or better delta values to get closer to the original image as searching for bettter matches takes time. Especially because it's lossy, there is no "correct" answer. There is just opinion.