Definition

What Is Low Latency? Latency Explained for Chat, Calls and Streaming

A fast connection can still feel slow. Here's what latency actually is, how much of it chat, calls and live streams can take, and where the milliseconds go.

Latency vs bandwidth

Latency

Bandwidth

What it measures

How long data takes to arrive

How much data can move per second

Unit

Milliseconds (ms)

Megabits per second (Mbps)

Analogy

The length of the road

The number of lanes

What users notice when it's poor

Lag, people talking over each other, slow replies

Low video quality, slow downloads, buffering

Main limits

Distance, hops, queues, processing

Your plan, the link and how many people share it

Can you buy more?

Only by getting closer or cutting steps

Yes, with a faster plan or link

Matters most for

Calls, games, live interaction, typing indicators

Video quality, file transfers, streaming at high resolution

What is latency?

Latency is the time it takes for data to get from one point to another. On a network it's quoted two ways: one-way latency, from sender to receiver, and round-trip time (RTT), there and back. The ping command shows RTT.

"Low latency" is relative to the job. A web page that loads in 300 ms feels fast. A voice call with 300 ms of one-way delay feels broken, because people start talking over each other. So the useful question is how much latency a specific experience can take, and how close you are to that limit.

Where latency comes from

Total delay is the sum of several smaller delays. For a voice or video call, the main ones are:

  • Capture and encoding. The app collects a frame of audio (often 20 ms) or video and compresses it.

  • Propagation. Light in fiber travels at about two-thirds of its speed in a vacuum, roughly 1 ms for every 200 km. A call between New York and London has a hard floor of about 28 ms each way.

  • Queuing and routing. Each router adds a little delay, and congested ones add a lot.

  • Jitter buffer. The receiver holds packets briefly to smooth out jitter. It's often the largest delay you control.

  • Decoding and playback. The device decompresses the media and sends it to the speaker or screen.

Add these up and you have a latency budget. If the total goes over the limit, something has to shrink.

Latency budgets: chat vs voice vs live streaming

Voice and video calls

ITU-T G.114 is the reference. It recommends keeping one-way delay under about 150 ms for most conversational use. Microsoft's Teams targets are tighter for the network part alone: under 50 ms one-way (100 ms RTT) from the user's device to Microsoft's edge.

Live streaming

Traditional HTTP streaming favors reliability over speed, with players buffering several seconds of video. Apple says Low-Latency HLS brings delay down into the range of standard television broadcasts. For two-way moments, such as a host taking questions on stage, apps use WebRTC instead, which runs at call-level latency.

Chat messages

Chat is the most forgiving. A message has to arrive complete and in order, so it travels over a reliable TCP connection such as a WebSocket. A delay of a few hundred milliseconds is barely noticeable. Latency matters more for small signals, such as typing indicators and presence, which feel wrong if they lag behind what's happening.

Rule of thumb: bandwidth decides how good a call looks. Latency decides whether it feels like a conversation.

Where low latency matters

  • In-app voice and video calls: telehealth visits, support calls, tutoring and dating apps.

  • Interactive live streams: live shopping, auctions, sports watch parties and webinars with Q&A.

  • Real-time chat: messages, reactions, typing indicators and read receipts that update as they happen.

  • Multiplayer games and remote control: every input needs a fast response.

  • Financial trading: where milliseconds affect prices.

How to reduce latency, and the trade-offs

  • Use servers close to your users. Distance is the one delay you can't optimize away. Pick the nearest region or edge.

  • Keep connections open. A persistent WebSocket avoids a new TCP and TLS handshake for every message. TLS 1.3 also needs one fewer round trip than TLS 1.2 to set up.

  • Use UDP for media. TCP retransmits lost packets and holds everything behind them. For calls, a late packet is as bad as a lost one.

  • Avoid VPN hairpins. Routing call traffic through a distant VPN gateway adds hops and delay.

  • Watch your buffers. Bigger jitter buffers and longer stream buffers make playback smoother but add delay. Lower latency always costs some resilience.

  • Measure the right thing. Ping shows RTT to one server. For calls, use roundTripTime from WebRTC stats on the actual media path.

Low latency in CometChat chat and calls

If you're adding chat or calling to your app, the two features have different latency needs, and CometChat handles them with different transports.

Persistent connections for chat, WebRTC for calls

The CometChat SDKs keep a WebSocket connection open to deliver messages, typing indicators and receipts in real time. You initialize the SDK with your App ID and region. CometChat Calls is built on WebRTC and the docs describe low-latency audio streaming, adaptive bitrate and an SFU for group calls, where each participant uploads one stream. Media prefers UDP and falls back to TCP on port 443 through TURN when UDP is blocked.

Explore CometChat Calls · Start building free

Frequently asked questions

What is considered low latency?

It depends on the use. For voice and video calls, under 150 ms one-way. For online games, often under 50 ms round trip. For live streaming, a few seconds counts as low compared with the 10 to 30 seconds of older setups.

Is low latency good or bad?

Good. Lower latency means less delay between an action and its result. The only cost is that systems with very little buffering have less room to absorb network problems.

What is the difference between latency and bandwidth?

Bandwidth is how much data a connection can carry per second. Latency is how long each piece of data takes to arrive. A connection can have plenty of bandwidth and still feel slow if latency is high.

What is a good ping for video calls?

A round-trip time under 100 ms to the media server is good. Microsoft's Teams target is under 100 ms RTT from the user's device to its network edge.

Does low latency use more data?

Not directly. Low latency comes from shorter paths and smaller buffers, not more data. Some techniques that protect low-latency calls, such as error correction, do add a little overhead.

Related terms