A keyframe is the everyday term for an I-frame — a video frame that's stored as a complete picture rather than as a small change from neighbouring frames. The word "keyframe" comes from animation tradition (the key drawings between which animators interpolate) and got reused for video. In a streaming context, keyframes are the points at which a viewer can start watching, seek, or switch quality renditions.
The reason keyframes have their own everyday name is that they're the most commonly visible unit of video structure for users and product teams. When a player's seek bar shows tiny preview thumbnails, those thumbnails are typically extracted at keyframe positions. When a user scrubs through a video, the player jumps to the nearest keyframe and starts decoding from there. When you ask FFmpeg to "split this video every 2 seconds for HLS", you're asking it to align segment boundaries with keyframes.
For a product team, two practical things to know. First, "keyframe interval" is the single setting that controls how often I-frames appear, typically expressed as seconds or as frame count. 2 seconds is the everyday VOD default for streaming. 1 second or less for low-latency live. 5+ seconds only if seek and rendition switching aren't priorities. Second, "keyframe" in casual conversation is ambiguous between a regular I-frame and the stricter idr (Instantaneous Decoder Refresh). For adaptive streaming, you almost always want IDR keyframes specifically — not any I-frame — to ensure clean segment boundaries.
