Latency

How far behind real time a stream runs — the difference between a webinar that feels live and one where the chat is 30 seconds ahead of the video.

Updated: May 2026 5 min read Streaming
In short

Latency is the delay between a moment happening live and a viewer seeing it on screen; standard HLS runs 10–30 seconds behind, while Low-Latency HLS cuts that to a few seconds.

What is latency?

Latency is the delay between a moment happening in front of the camera and a viewer seeing it on screen. In a live streamLive streamReal-time video delivered as it is captured, transcoded and segmented on the fly — webinars, events, broadcasts.Open definition →, it's the gap between real time and playback — measured in seconds.

For on-demand video it barely matters; a VoDVoD (Video on Demand)Pre-recorded video a viewer can start any time — the opposite of live. Transcoded once, then served from the CDN on request.Open definition → file is already complete, so there's nothing to be "behind". For live it can make or break the experience: if viewers interact through chat, polls or bidding, a 30-second lag means the conversation runs ahead of the video.

Latency is not the same as bufferingBufferingWhen playback pauses to load more data. ABR exists largely to prevent it by stepping quality down before the buffer empties.Open definition →. Buffering is playback pausing to load data; latency is how far behind real time you are. A stream can be perfectly smooth and still 20 seconds late.

The latency tiers

"Low latency" means different things depending on the delivery method. The rough tiers:

TierTypical delayMethodGood for
Standard HLS10–30 sRegular segmentsBroadcasts, one-way events
Low-Latency HLS 2–8 sPartial segments + server pushWebinars, Q&As, interactive live
WebRTCUnder 1 sPeer real-time transportVideo calls, two-way conversation

Most interactive live sits comfortably in the LL-HLSHLS (HTTP Live Streaming)Apple’s adaptive streaming protocol. Uses .m3u8 manifests and segmented delivery; supported on every browser, device and smart TV.Open definition → range — a few seconds — which balances responsiveness with the reliability of HTTP delivery.

What adds latency

Encoding

Compressing the live feed into deliverable renditions takes time before anything can ship.

Segmenting

The stream is cut into segments; longer segments are more stable but add delay before each is ready.

Player buffer

Players hold a few segments in reserve before playing to avoid stalls — reserve that also adds latency.

Low-Latency HLS

LL-HLS uses partial segments and server push so the player can start a chunk before it's fully written, cutting the delay to a few seconds.

When you need low latency

Chasing the lowest possible latency has a cost: smaller segments and a shallower buffer leave less headroom to absorb a bandwidth dip, so an aggressive low-latency setup can rebuffer more. So match the target to the use case rather than defaulting to "as low as possible":

Interactive live → seconds

Webinars, Q&As, live shopping and auctions need a few seconds so chat and reactions stay in sync. LL-HLS is the sweet spot.

One-way broadcasts → tens of seconds is fine

Keynotes, conferences and passive viewing don't need real-time; standard HLS buys more stability at scale.

On-demand → latency is irrelevant

A recorded replay has no real-time reference at all — only start time and smoothness matter.

Further reading

Denis Konnov

Written by

Denis Konnov Verified

CMO at Kinescope

Denis leads marketing and content at Kinescope, working closely with the platform's engineers. This glossary is reviewed and kept up to date by the team building Kinescope's streaming, encoding and CDN delivery infrastructure.

FAQ

Can’t find the answer you need? Check out our Help Center.

Latency is the delay between something happening in front of the camera and a viewer seeing it. Standard HLS live typically runs 10–30 seconds behind; Low-Latency HLS brings that down to a few seconds, which matters for anything interactive like webinars, auctions or Q&As.

Latency is how far behind real time the stream is; buffering is playback pausing to load more data. You can have low latency and still buffer on a weak connection, or a rock-solid stream that is simply 20 seconds behind — they are separate issues.

Low-latency streaming reduces the delay to a few seconds using techniques like Low-Latency HLS (partial segments and server push). It is needed when viewers interact in real time — live chat, polls, bidding — where a 30-second lag would break the experience.

Several stages add up: encoding the stream, cutting it into segments, the player buffering a few segments before playback, and network travel time. Longer segments and bigger buffers add latency but improve stability, so it is always a trade-off between delay and smoothness.

No — it is a trade-off. Cutting latency means smaller segments and a shallower player buffer, which leaves less headroom to ride out a bandwidth dip, so aggressive low-latency setups can rebuffer more on weak connections. Match the target to the use case: interactive live needs a few seconds; a broadcast or replay does not.

Kinescope live streaming runs at roughly 5–10 seconds of signal delay, with Low-Latency HLS available for lower-latency, interactive broadcasts such as webinars and Q&As. Recording to on-demand is automatic when the stream ends.