Transcoding is converting a video from one form to another — most often "the file we have" into "the files we need". A user uploads a 2 GB MOV from their phone; the platform transcodes it into a 50 MB H.264 stream for old browsers, a 30 MB HEVC stream for iPhones, an 18 MB AV1 stream for modern devices, and three smaller renditions for slow networks. One upload becomes six output files, each tuned for a different audience.
Why this matters: the source video that arrives from a camera or editor is almost never the right format for delivery. It's too big, in the wrong container, in the wrong codec, at the wrong resolution, or with the wrong rate control mode for streaming. Transcoding bridges the gap, and it's the most common operation in any video platform — YouTube transcodes every upload roughly 20 times to populate its adaptive bitrate ladder; Zoom transcodes thousands of live participants per second.
Two things to distinguish. Transcoding changes the codec or quality (decode → re-encode); it costs CPU/GPU time proportional to video length and is the heavy work. Transmuxing changes only the container (e.g. MP4 to MPEG-TS); it's nearly free because no re-encoding happens — the codec data is just repacked. A typical production pipeline uses both: transcode once into multiple quality renditions, then transmux on the fly into whichever container a specific player wants. Cloud services like AWS MediaConvert and Bitmovin are essentially managed transcoding-at-scale.
