Hybrid codec is the architectural pattern that every practical video codec since the 1980s has followed. It combines two fundamentally different compression techniques into one pipeline: prediction (use what we already decoded to guess what comes next) and transform coding (express the small leftover differences as frequency-domain coefficients that compress efficiently). Every modern codec — MPEG-2, H.264, HEVC, VP9, AV1, VVC — is a refinement of this same hybrid template; the differences between them are quantitative, not architectural.
The flow inside a hybrid codec: split the frame into blocks; for each block, predict its content from already-decoded blocks in the same frame (intra-prediction) or from earlier frames (inter-prediction); subtract the prediction from the real pixels to get a small residual; transform that residual with a dct or similar; quantize the result aggressively; entropy-code the final numbers. Decoder runs the same steps in reverse. Every box in this pipeline has been refined over 40 years of research, but the overall shape hasn't changed since H.261 in 1988.
For a product team, "hybrid codec" is the mental model that makes codec discussions tractable. New codec X says it's "30 % more efficient than its predecessor"? That gain came from somewhere in this pipeline — usually better prediction (more directional modes, more reference frames, sub-pixel precision), more flexible block partitioning, smarter transform menus (ADST, integer transforms), more sophisticated rate control, better entropy coding (CABAC over Huffman, arithmetic coding over CABAC). Knowing the hybrid template makes it easy to mentally place any new claim into context: which box of the pipeline did this improve?
