Does it have to know the information for all frames, or just the key frames? If it’s the later, then you could encode so the initial table, has X key frames spr

Does it have to know the information for all frames, or just the key frames? If it’s the later, then you could encode so the initial table, has X key frames spread out over a few hours, then force a frame to exist in those locations when encoding the live stream.