Transcoding is decoding a compressed video and encoding it again. Any change to codec, resolution, frame rate or bitrate requires it. It is distinct from remuxing, which repackages the same encoded streams into a different container and costs nothing in quality.
The part people underestimate is how often it happens without their involvement. Upload a video to Instagram, TikTok, WhatsApp, YouTube or almost any other platform and it will be transcoded on arrival to the platform's own specification. Your file is an input to their encoder, not the thing viewers see.
Three practical conclusions:
- Uploading a larger file does not produce a better result beyond the point where the platform's encoder is the limiting factor. Past that, the extra data is discarded on their servers.
- Leave headroom before an upload that will be transcoded. Compressing to the visible edge of acceptable and then handing the result to a second lossy encoder is how detail disappears. A file that scores 95 VMAF going in does not come out at 95.
- Match the platform's expected resolution and frame rate where you can. A transcode that also has to rescale does more damage than one that does not.
The corollary for your own workflow: compress once, from the original, as late as possible.