RTMP was created by Macromedia (later Adobe) for the Flash Player. It carries H.264 video and AAC audio in chunks over a long-lived TCP connection, supports a simple key-based authentication, and is trivial to encode and decode. When Flash died, RTMP outlived it because every encoder, every live platform and every OBS user already spoke it. Twitch, YouTube Live, Facebook Live and every smaller live service still accept RTMP ingest in 2026, even when they internally repackage to HLS or WebRTC for distribution.
RTMP's weaknesses are: TCP-based (so packet loss inflates latency through retransmission), single video and single audio per session (no native simulcast or alternate-language tracks), no built-in encryption (RTMPS adds TLS but is not universally supported on hardware encoders), and a fixed 1-second timestamp granularity that complicates frame-accurate sync. End-to-end latency over RTMP is typically 2–5 seconds — fine for casual live, not for low-latency.
The successor pattern is to keep RTMP as a fallback ingest while encouraging encoders to switch to SRT for unicast contribution or WHIP for browser/software contribution. SRT gives sub-second latency with adjustable loss recovery; WHIP plugs into WebRTC for sub-second ingest from any browser. Most managed services accept all three on the same URL pattern and let the encoder choose.
