encoding a video to make it easier to datamosh

so instead of it being an accident, I want to intentionally recreate this. Like I said, this is experimental, but it does in fact work with avi files, and in 'real life'(like on tv or something - natural error) when an i-frame is somehow missing in an mpeg file, it should have the same effect.

Though the file that is produced is not readable by all players, the players that do read it show strange pixelated corruption that appears as a result of these references being missing/wrong. Essentially, what I successfuly did with avi files and would like to do with the mpeg files is in fact remove the i-frames and intentionally create a broken file, with missing information. And I am well aware of how the compression works by having i-frames that hold all information of the one image, while P and B frames save file space by only holding reference information that is based on the i-frame. Your knowledge is correct with regards to how mp4 and mpeg2 work.

If you remove the I frame there is nothing to reference subsequent frames to. An I frame is sent that is the complete frame, subsequent intermediate frames are simply a code saying, the same as the last frame except for this bit, this bit, and this bit, then when the next I frame comes along, it can compare the guesses made on previous frames and confirm they are correct. I may be completely wrong as I never work with mp4 files, only DV-AVI and DVD compliant mpeg2, but will you even have anything at all if you remove the I frames? Assuming mp4 is a better compression than mpeg2 but works in the same way, the I frames are the only ones that contain the full frame.

Encoding a video to make it easier to datamosh