What's the difference between Frame and GOP(K-frame)?

[immortal192]

I'm not looking to be an expert video editor--I just needed something that can trim/join files without losing any noticeable quality. Someone recommended Solveigmm Video Splitter so I am using that at the moment. In the settings, it shows 2 trimming accuracy options: Frame and GOP(K-frame).

If I'm not mistaken, Solveigmm advertises the use of GOP(K-frame) and not Frame, yet the default was set to Frame which I found bizarre. What is the difference between the two and which one is which? I believe one of them allows trimming with an accuracy of a single frame but the few frames where it was trimmed will be at ~99% quality (while the rest of the video is 100% quality) while the other won't be as accurate (may be off by a few frames but the entire video will be 100% quality (i.e. quality fully preserved).

And because this is "smart mode" it doesn't really matter which option to use as humans won't be able to tell the quality lost even with the method of ~99% quality for a few frames. Still, I am curious which one is which and if there are other factors I might want to consider when choosing one over the other. Also, is "smart mode" what the industry calls it or just Solveigmm? What is the generic name for such a feature?

Thanks!

[jagabo]

Your understanding of the situation is about right. GOP level accuracy means you can only cut at key frames but nothing has to be reencoded. Frame level accuracy means you can cut at any frame, and with the addition of smart rendering, only the cut GOPs will be reencoded. But it may be more than just a "few frames" that get reencoded. For example, x264's default GOP size is 250 frames. That's 10 seconds in a 25 fps video. Many h.264 capture devices use GOP sizes from 30 to 60. DVD uses about 15.

High compression codecs get a large part of their compression by not repeating parts of the picture that don't change from frame to frame. The first frame is encoded much like a JPEG image, the entire frame can be reconstructed from just that data. This is called a key frame. But subsequent frames only contain the changes from the last (or other) images. For example, the second frame will say "this frame is the same as the first one, just make these changes..." The third frame will say "this frame is the same as the second, just make these changes..." Eventually a new GOP will be started with a new key frame. To reconstruct the last frame of a GOP you have to go back to the previous key frame, then reconstruct every frame between that key frame and the last frame.

[immortal192]

Wow, I learned a lot. Thank you.