A live media streaming architecture with real-time live watermarking wherein watermarks are added at a frame or subframe level. The architecture can use a UDP/WebRTC architecture, without requiring a content delivery network, which is not dependent on transmitting media by segments, thus allowing insertion of the watermark at a frame or subframe level. By inserting watermarks at a frame or subframe level, latency is minimized for media delivery to under 500 milliseconds (ms) and for extraction to about five to fifteen seconds.