How to reduce live-stream latency (and measure it first)

If your “live” stream is 20 seconds behind reality, viewers notice — and for betting, auctions or live shopping, those seconds are money. Here’s what actually drives latency, and how to get it down to sub-second.

What “latency” means in live streaming

Latency in live streaming is the glass-to-glass delay: the time from something happening in front of the camera to a viewer seeing it on their screen. It’s not one number you set — it’s the sum of capture, encode, ingest, packaging, delivery and the player’s buffer. The player buffer is usually the biggest hidden chunk.

Why standard HLS is 15–30 seconds behind

Classic HLS (the default for most web video) chops the stream into segments of several seconds each and asks the player to buffer a few of them before playback. Two to three segments of 6–10 seconds each is already 15–30 seconds of delay before a single packet is “late.” It’s rock-solid and scales beautifully — it just was never designed to be fast.

The latency ladder

Lower latency is a spectrum, and each rung trades a little compatibility for a lot of speed:

The rule of thumb: if a viewer ever has to act on what they see — place a bet, bid, hit “buy” during a drop, ask a question — you need sub-second, and that means WebRTC. If they only watch, LL-HLS is plenty.

Step 1 — measure what you actually have

Before you change anything, find out how far behind live you really are — from a viewer’s seat, not your encoder’s dashboard. Encoders report the latency they add; they can’t see the buffer the player adds at the other end. The only honest number is measured end-to-end.

Step 2 — reduce it

Four levers, roughly in order of impact:

• Pick the right protocol. Moving from standard HLS to LL-HLS cuts you to a few seconds; moving to WebRTC gets you under one. This single choice dominates everything else.
• Put delivery near your viewers. A stream served from one region to a global audience adds real distance-latency and buffering for far-away viewers. Regional edges close to your audience remove that gap — this is where “my Asia viewers buffer” problems disappear.
• Right-size the bitrate ladder. A proper adaptive-bitrate ladder keeps weak connections stable at a lower rung instead of stalling, so “buffering” doesn’t masquerade as latency.
• Tune ingest. Low-latency ingest (SRT, WHIP) from the encoder keeps the front of the pipeline tight so your delivery gains aren’t eaten before the stream even arrives.

When sub-second is worth it

Sub-second delivery isn’t free complexity for its own sake — it’s the difference between a product that works and one that doesn’t, in exactly these cases: live casino and betting (a bet on a stale picture is a lost bet), auctions (every bidder must see the same frame), live commerce (the “add to cart” has to land during the hype, not after), and interactive events (Q&A falls apart at 20-second lag). For everything else, a few seconds is fine.