Low Latency Streaming: Protocols, Settings and Network Fixes

Low latency streaming means keeping the delay between your camera and your viewer's screen short enough that the broadcast feels live. The usual yardstick is glass-to-glass delay: the time from an event happening to it playing on a viewer's device, as RFC 9317 defines it. How low you should go depends on what you stream. A chatty Twitch show and a live auction need very different targets, and the lowest setting is often the wrong one.

This guide covers what the latency tiers mean, which protocols deliver them, which encoder settings matter, and why your network, not your plan's headline speed, usually decides whether low latency holds up.

Grandspeed

Check the Connection Behind Your Stream

Test upload speed, latency, jitter, packet loss, and bufferbloat to see more of the network conditions affecting your stream.

Test your connection

TL;DR

Industry surveys group live latency into ultra-low (under 1 second), low (1 to 10 seconds) and traditional (over 10 seconds), according to a 2023 survey of live streaming evolution.
WebRTC is the main protocol for sub-second delivery. Most other live broadcasts are well served by low-latency modes in the 2 to 10 second range.
Lower latency shrinks the player's buffer, so viewers feel every jitter spike and congestion dip more.
Use CBR encoding and a 2-second keyframe interval, and keep upload bitrate well under what your connection sustains.
Test upload stability under streaming load, not just peak speed. Our livestream monitor is built for that.

What low latency actually means
Which latency level do you need?
Low latency protocols compared
What the big platforms offer
The trade-off nobody mentions in the settings menu
Encoder settings that help
Why a fast plan does not guarantee low latency
How to tell whether your network is the problem
A worked example
Our perspective
Test your connection before you go live
FAQ
Sources

What low latency actually means

There is no single definition of "live". Dolby's guide notes that latency varies widely with technology and configuration, and different vendors draw the line in different places. Wowza describes low latency as roughly five seconds or less, while Mux puts it at around five to twelve seconds. Amazon's IVS documentation separates low latency (under 5 seconds) from real-time (under 300 milliseconds).

The academic grouping is a useful common frame, per the survey linked above:

Tier	Glass-to-glass delay	Typical use
Ultra-low	Under 1 second	Live wagering, auctions, interactive sports
Low latency	1 to 10 seconds	Streamer chat, live events, general interactive broadcasts
Traditional	Over 10 seconds	Linear-style broadcasts where quality and reach matter most

Two terms get confused here. Latency is the delay in the video pipeline end to end. Network latency, or ping, is the round-trip time of packets between two points. Ping is one small piece of glass-to-glass delay. Encoding, packaging, CDN delivery and the player's buffer usually add far more.

Which latency level do you need?

Most streamers should pick the slowest setting that still serves their audience. That sounds backwards, but the reasoning is solid. RFC 9317 on streaming operations explains that pushing to sub-second delivery on ordinary consumer internet tends to produce frequent visible artifacts, because the target becomes as small as normal jitter and network fluctuation. The practical advice from that document is to match the latency mode to the use case, such as betting versus news, rather than defaulting to the fastest option.

A rough decision guide:

Real-time interaction is the product (auctions, betting, remote guests, live coaching): you need ultra-low latency and should plan the whole pipeline around it, often in a controlled environment.
Live chat is part of the show (gaming, Q&A, IRL): low latency in the 2 to 10 second range keeps conversation natural without starving the buffer.
One-way broadcast (conference talk, sermon, concert): normal latency gives the best quality and the most resilience.

Consistency matters too. Dolby's streaming solutions page points out that a steady delay can be more valuable than the absolute minimum when video must stay in sync with other live data, such as odds or scores.

Low latency protocols compared

The protocol you use to ingest the stream and the one used to deliver it are often different. Here are the main ones in plain terms.

WebRTC

WebRTC is the primary choice for sub-second delivery, according to Dolby's low latency guide. It is built for browser-based real-time communication, which is why video calls and interactive platforms use it. The trade-off is scale and complexity: it needs more server resources per viewer than segment-based delivery, and it gives you less room to absorb a bad network.

Low-Latency HLS and LL-DASH

These extend the HTTP-based formats that run most of the internet's video, using smaller chunks so the player can start earlier. They sit in the low-latency tier, typically a few seconds, and work with standard CDNs. For broad audiences watching on every device, they are usually the sweet spot.

SRT and RTMP for ingest

RTMP is still the most common way encoders such as OBS push a stream to a platform. SRT is a newer transport designed to handle packet loss over unreliable networks, and it is popular for contribution links from the encoder to a media server. Neither is a viewer-side format, but both affect how much delay you add before the stream even reaches the CDN.

What to take from this

The protocol sets the floor. Your encoder settings, your upload connection and the viewer's player set how close you actually get to it.

What the big platforms offer

YouTube exposes three modes in its live streaming latency help page: normal, low and ultra-low. Ultra-low latency, typically under about five seconds, enables near real-time conversation but does not support 4K. Low latency, under about ten seconds, balances interactivity and playback stability. Normal latency is tuned for quality and supports every resolution but does not suit real-time chat.

Twitch has its own low latency setting, documented in the Twitch Help Center. You find it in your dashboard under the Stream preferences. If your channel is chat-driven, it is worth testing. If your connection is shaky, you may prefer to switch it off.

If you stream to Twitch and notice lag on either side of the screen, our Twitch stream lag guide walks through what to check first. For YouTube, the two-minute buffering checklist covers both streamer and viewer sides.

What the big platforms offer — overview diagram

Lower latency means a smaller read-ahead buffer in the viewer's player. That is what makes interaction possible, and it is also what makes playback more fragile. With less buffered video in reserve, a brief congestion dip between you and the viewer is more likely to turn into a visible stall. YouTube's help documentation says as much when it describes the trade-off between interactivity and buffering risk.

Smaller video buffer stalls during congestion

So low latency does not make a stream better. It makes it more interactive and less forgiving. That is why the network underneath it matters so much.

Encoder settings that help

These are the settings professionals reliably recommend, and none of them cost anything.

Use CBR (constant bitrate). Predictable output is easier for ingest servers and players to handle. Dolby recommends CBR for low latency work.
Set a 2-second keyframe interval. YouTube's encoder settings guidance asks for this, and Twitch workflows use the same convention. Keyframes are where players can join and recover, so a steady interval keeps startup and recovery predictable.
Choose a bitrate your upload can sustain. YouTube's own recommendation for 1080p at 60 frames per second with H.264 is around 12 Mbps. That figure only works if your upload holds it without dips.
Use wired Ethernet for the streaming PC. Wi-Fi adds jitter, which interacts badly with small buffers.
Keep CPU and GPU headroom. Overloaded encoders cause problems before the network is even involved.

If you are unsure what bitrate fits your resolution and frame rate, our video bitrate calculator gives you a number to start from, and our bitrate versus resolution guide offers three copy-ready presets with a 1.5x headroom rule.

Why a fast plan does not guarantee low latency

You can pay for a gigabit plan and still have a stream that stutters. Raw speed measures how much data moves per second. Low latency streaming depends on how consistently it arrives.

Three network problems cause most trouble:

Jitter: variation in packet arrival times. Small buffers cannot absorb it.
Packet loss: packets that never arrive and must be resent or skipped.
Bufferbloat: delay that builds when your router's queues fill under load, often when something else on the network is uploading.

RFC 9317 notes that effects like bufferbloat and packet reordering happen on the same time scale as aggressive latency targets, which is exactly why sub-second goals are hard on the open internet.

A standard speed test tends to miss these. It reports a peak number from a short burst. We cover this gap in speed test alternatives that catch jitter and packet loss.

How to tell whether your network is the problem

Watch the encoder's own stats

In OBS, the stats window shows dropped frames. Per an OBS Forums explanation, dropped frames are encoded frames that could not be sent because of network issues. Frames skipped due to CPU or GPU overload happen earlier and are a different problem. If the dropped count climbs while your bitrate wobbles, look at your upload connection first. The same forum thread mentions enabling dynamic bitrate as a quick fix.

Test under streaming conditions

A 10-second test while the household is idle tells you little. A better routine:

Run the test at the time of day you usually stream.
Close other uploads, cloud backups and game downloads, or deliberately leave them running if that mirrors your real evening.
Look at upload speed, jitter, packet loss and latency under load, not just the headline number.
Repeat on different days. Upload speed often swings; our guide on why upload speed fluctuates explains common causes.
Compare your sustained upload with your target bitrate. We suggest roughly 1.5 to 2 times headroom, which is the thinking behind our best internet for streaming guide.

If the upload number is lower than you expected, see why your upload speed may be so slow for causes and fixes.

A worked example

Say you stream 1080p60 to YouTube at 12 Mbps with ultra-low latency selected. Your plan advertises 20 Mbps upload. On paper, that looks fine. In practice, your sustained upload during the evening may dip to 14 Mbps with jitter spikes whenever a family member uploads a video. At 12 Mbps you have under 20 percent headroom, and the viewer's buffer is tiny.

You have three sensible options:

Drop to a lower bitrate or 720p60 and keep ultra-low latency.
Move from ultra-low to low latency and give the player more buffer.
Fix the cause, for example by wiring the PC or limiting background uploads on the router, then retest.

The numbers here are illustrative, but the logic holds: stability under load decides which latency mode is realistic.

Our perspective

We built GrandSpeed because a single download figure does not explain why a stream stutters. In our experience with the kinds of problems streamers bring us, the culprit is rarely that the plan is too slow. It is that upload stability, jitter, packet loss or bufferbloat are not what the plan's headline suggests, and nobody measured them under streaming load.

That is why we push people to treat latency mode as a decision, not a badge. If your connection cannot sustain 1.5 times your bitrate with steady jitter, ultra-low latency will cost you more in artifacts and buffering than it earns in chat speed. Measure first, pick the mode second, and retest when something in your setup changes.

— Peter D

Test your connection before you go live

Our livestream upload speed monitor checks how your connection holds up under streaming conditions, and the video streaming quality test shows which resolutions your tested speeds support. For a quick baseline, run the GrandSpeed speed test, which also measures jitter, packet loss, bufferbloat and DNS resolution time. Everything is free and works in any modern browser on phone, tablet or laptop. For a fuller walkthrough of the setup side, see our livestreaming guide.

FAQ

Is low latency mode good for streaming?

It is good when viewers interact with you in real time and your upload is stable. The cost is a smaller playback buffer, so viewers with weak connections buffer more often. If your audience just watches, normal latency usually gives a smoother picture.

What is good latency for streaming?

For chat-driven streams, a few seconds up to about ten is a reasonable target, and it falls in the low-latency tier. Sub-second delivery is for interactive products such as auctions or betting. Pick the figure that matches the experience you want, then confirm your network can support it.

What is ultra-low latency streaming?

It is delivery with under one second of glass-to-glass delay, usually built on WebRTC. Platforms define it differently: YouTube's ultra-low setting is typically under about five seconds, and it drops 4K support. Check the definition each platform uses before you compare options.

Does a faster internet plan reduce stream latency?

Not by itself. Latency depends on the protocol, encoder settings, distance to the server and network stability. Extra bandwidth helps only if you were running short of upload headroom, and jitter or bufferbloat can cause problems on any plan.