Internet Calling Explained: How VoIP and WebRTC Actually Work

An internet call is a phone call where your voice travels as data packets over the internet instead of over the telephone network. That one sentence explains almost everything downstream: why internet calling is cheap or free, why quality varies with your Wi-Fi, why a web call needs no phone line, and why calling a real phone number still costs money even when the internet part is free.

This is the explainer for people who want to actually understand it. Not to pass an exam, but because knowing what happens under the hood makes you better at choosing services, diagnosing bad audio, and spotting nonsense in marketing copy.

The vocabulary, in one table

Eight terms carry this whole topic. Every one of them gets used properly further down, but here they are in plain words, for reference and for decoding provider marketing:

Term	What it means
VoIP	Voice over IP: your voice carried as internet data packets instead of over a dedicated phone line.
PSTN	The public switched telephone network, the traditional global phone system that every real phone number lives on.
SIP	The protocol that sets internet calls up and tears them down: who's calling whom, ringing, answering, hangup.
WebRTC	The calling engine built into every modern browser, which is why a website can place a call with nothing installed.
Codec	The compressor that turns your voice into a small data stream and back; Opus is the modern standard.
Jitter	Unevenness in how packets arrive; the main cause of robotic, garbled call audio.
Latency	How long your voice takes to reach the other ear; too much of it makes people talk over each other.
Termination fee	The per-minute charge a carrier collects for completing a call to a phone number; the reason calls to real numbers are never free and unlimited at once.

VoIP vs PSTN: two networks, one conversation

The traditional phone system is the PSTN, the public switched telephone network. Its defining idea is the dedicated circuit: when a call connects, the network reserves a path between the two phones for the whole conversation, and that path carries your voice and nothing else until you hang up. It's reliable and it's why landline audio always sounded consistent. It's also expensive, because reserved capacity sits there whether you're talking or breathing.

VoIP, voice over IP, throws out the reserved circuit. Your voice becomes small data packets that share the internet with everyone's video streams and downloads, each packet finding its own way to the destination. No reservation, no dedicated line, no per-call infrastructure. The economics follow directly: carrying a voice call over the internet costs the provider roughly what carrying any small data stream costs, which is close to nothing.

The two networks still have to interoperate, because a billion phone numbers live on the PSTN side. Every service that lets an internet call reach a real phone number operates or rents gateways where the two worlds meet, and that gateway crossing is where the money is. More on that below.

What actually happens when you make an internet call

Take a simple case: you open a web dialer, type a number, and press call. Here's the journey, stage by stage.

Your voice becomes numbers. The microphone captures sound as an analog wave, and your device samples it thousands of times per second into digital values. A codec (coder-decoder) then compresses those values hard. Modern voice codecs like Opus get natural-sounding speech into somewhere around 20 to 40 kilobits per second, a sliver of any broadband connection.
The numbers become packets. The compressed audio is chopped into packets, each carrying roughly 20 milliseconds of sound plus addressing information and a sequence number. Fifty times a second, a packet leaves your machine. Each one is encrypted before it goes.
Signaling sets up the call. Separately from the audio, the service needs to establish who is calling whom, negotiate which codec both ends support, and handle ringing, answering, and hangup. The classic protocol for this is SIP, the session initiation protocol, which is essentially the internet's replacement for a dial tone. Browser-based services do the same job over their own connections, but the choreography is the same: invite, ring, answer, talk, bye.
Media servers relay the audio. Packets usually don't flow directly between the two ends. They pass through the provider's media servers, which relay them through firewalls, mix group calls, and bridge between different networks. Each hop adds a little delay, which is why providers scatter servers around the world to keep the path short.
The gateway crosses into the phone network. If you dialed a real phone number, the provider's PSTN gateway takes your packets, converts the audio into the phone network's format, and hands the call to a carrier that completes it to the destination number. The receiving carrier charges a termination fee for that, a fraction of a cent to a few cents per minute depending on country and whether it's a landline or mobile. This fee is the entire reason "free calls to real numbers" is always capped somewhere, a pattern you can see across every service in the roundup of free ways to call online.
The far end reverses everything. Packets are reordered by sequence number, decoded back into audio, and played. Total elapsed time from your mouth to their ear, on a good connection: one to three tenths of a second.

Latency and jitter, in plain words

Two words explain nearly every quality complaint about internet calls.

Latency is how long a packet takes to arrive. Below about 150 milliseconds one way, conversation feels natural. Beyond 300, you get the talking-over-each-other effect familiar from old international calls: both people start speaking, both stop, both apologize. Latency comes from distance, from the number of server hops, and from congested links along the way.

Jitter is the variation in latency, and it's the more vicious of the two. Voice packets are sent at a perfectly steady rhythm, but the internet delivers them unevenly: three arrive at once, then a gap, then two more. The receiving end keeps a small buffer that smooths this out by holding packets briefly before playing them. When jitter exceeds what the buffer can absorb, packets are effectively late or lost, and the audio does the robotic garble everyone recognizes. This is why a call can break up on a connection that speed-tests beautifully. Speed tests measure throughput, and voice barely needs any; what voice needs is steadiness, and no speed test shows you that.

The practical consequences are simple. Wired beats Wi-Fi because radio interference is a jitter machine. Wi-Fi beats mobile hotspots for the same reason. And a household member starting a big upload can wreck your call while barely denting your bandwidth, because the upload fills queues in your router and jitter spikes.

Why quality varies between services as much as connections

Two services on the same connection can sound noticeably different, and the reasons are concrete: which codec they negotiate (Opus degrades gracefully, older codecs fall apart), how close their nearest media server is to you, how well their software conceals lost packets by interpolating over the gaps, and how aggressively their echo cancellation and noise suppression are tuned. None of this appears on a pricing page, which is why the only real test of a calling service is a call on your own network.

Why call quality varies (and how to fix it)

Everything above translates into a short diagnostic list, because bad internet-call audio almost always has one of four causes, and each has a fix that costs little or nothing.

The connection is jittery, not slow. This is the most common misdiagnosis. Voice needs almost no bandwidth, so upgrading your internet plan rarely helps; what breaks calls is packets arriving unevenly. The signature is audio that robotizes or drops syllables while web pages still load fine. The fix is stability: move closer to the router, get other heavy uploads off the network during calls, or best of all, plug in a cable.
You're on Wi-Fi when you could be wired. Radio is inherently uneven: interference from neighbors' networks, microwaves, and walls all produce exactly the jitter described above. A cheap ethernet cable (plus a USB adapter if your laptop lacks a port) removes the entire category of problem, and it's the single highest-value fix on this list for anyone who calls from the same desk every day.
You're using laptop speakers instead of a headset. With speakers, your microphone hears the other person's voice and sends it back to them as echo, and the echo cancellation that fights this also clips and distorts both ends. Any headset, even phone earbuds with an inline mic, breaks the loop and removes fan and keyboard noise as a bonus. If people say you sound distant or echoey, this is the fix, not a new service.
The browser or OS is squeezing the microphone. Silence or one-way audio is nearly always permissions: the site's mic permission denied in the browser, the browser itself blocked in the operating system's privacy settings, or the wrong input device selected when you have several. Check those three layers in that order before blaming the provider, because the same misconfiguration will follow you to any other service you try.

Only after those four does it make sense to suspect the service itself, and at that point the test is simple: same computer, same network, different provider. If the problem moves with the service, switch; if it stays, it's yours.

WebRTC vs installed softphones

There are two main ways internet calling reaches you as a user, and the difference is about packaging, not audio.

WebRTC is the calling machinery built directly into every modern browser: microphone capture, Opus encoding, encryption, and packet transport, all standard equipment since roughly 2013. It's what lets a website place a real call with zero installation, and it's the technology behind Google Meet, FaceTime links, and every web dialer. The full walkthrough of using it is in how to call from your browser.

Softphones are installed apps like Zoiper or Linphone that speak SIP directly to a VoIP provider you configure with your own credentials. They predate WebRTC, and they still win for heavy daily use: they run in the background to receive calls, integrate with headset buttons and system notifications, and give you provider choice instead of locking you to one website.

Under the hood, both end up doing the same thing: codec, packets, encrypted transport, gateway. Pick by usage pattern, not technology. Occasional calls favor the browser's zero setup; daily calling favors an installed app, and the comparison of the best apps to call from your computer covers the specific options.

A short history, including Skype's death

Internet calling is older than most people think. VoIP experiments date to the 1970s, and commercial "internet phone" software appeared in the mid-1990s, when calls sounded like walkie-talkies underwater and international phone rates were extortionate enough that people tolerated it.

Skype, launched in 2003, made it mainstream. Its peer-to-peer design delivered surprisingly good free calls between computers, and cheap SkypeOut calls to real phones undercut carriers by an order of magnitude. For most of two decades, "internet call" and "Skype" were synonyms. Microsoft bought it in 2011, then slowly starved it while building Teams, and in May 2025 retired Skype entirely, taking Skype-to-phone calling and Skype credit with it.

Skype's shutdown mattered less than it would have a decade earlier, because its ideas had already dissolved into everything else. WhatsApp normalized free app-to-app calls for billions. WebRTC put the calling engine into the browser itself, so a "web call" needs no product at all, just a tab. And the pandemic made talking to a computer screen unremarkable. Internet calling didn't decline with Skype; it became infrastructure, which is the more thorough kind of victory.

Frequently asked questions

Is an internet call the same as a VoIP call? Yes, effectively. VoIP is the technical term for voice carried over the internet, and internet calls, web calls, and WiFi calls are all VoIP under the hood. The label varies with the packaging: a browser tab, an app, or your carrier routing calls over Wi-Fi.

Why do internet calls sound robotic sometimes? Jitter: packets arriving unevenly enough that some miss their playback moment, forcing the codec to guess at the missing audio. It correlates with unstable connections rather than slow ones. Moving from Wi-Fi to a wired connection, or closer to the router, fixes it more often than upgrading your plan.

How much data does an internet call use? Very little. Voice runs around 0.3 to 0.5 megabytes per minute with a modern codec, so an hour of talking costs roughly the data of a couple of minutes of video streaming. Data volume is almost never the problem; connection stability is.

Can an internet call reach a normal phone number? Yes, through a provider operating a PSTN gateway, which converts the call and pays the receiving carrier a per-minute termination fee. That fee is why number-reaching services charge or cap you while app-to-app calling is free and unlimited.

What replaced Skype for internet calls? No single product. WhatsApp and FaceTime took personal calls, Teams and Zoom took work calls, and browser-based dialers took the dial-a-real-number case. If a guide still recommends Skype, it predates May 2025 and likely has other stale advice too.

Internet calling is one of the quiet triumphs of the internet: a service that used to cost dollars per minute now rides for free on infrastructure you already pay for. Once the concepts here click, the practical side follows easily, starting with the full map of ways to call from your computer.