How much data a video call actually uses

The first time I paid attention to any of this, I had three days left on a billing cycle, about four hundred megabytes of allowance, and a person on the other end of the call in the middle of a long story about her brother's wedding. I spent the second half of that conversation doing arithmetic instead of listening, which is a stupid way to spend a call. Afterwards I did the thing I should have done years earlier: opened the data screen on my phone, wrote the number down, made a call, and looked again. Then I did it the next day, and the day after that, for about a month.

None of this is laboratory work and I would not want it presented as such. It is one phone, one plan, one person's habits and a notes app. Before each call I reset the per-app statistics in settings, and when the call ended I wrote down the minutes and the megabytes. I did it on Wi-Fi and on mobile data, in a bright room and a dark one, on calls where I sat still and on calls where I wandered around the flat with the phone in one hand. The aim was never to publish a benchmark. The aim was to stop guessing, because guessing is what makes people ration a conversation they are enjoying.

Here is the headline. A one-to-one video call with both cameras on, at whatever quality my phone settled into by itself, cost somewhere between six and nine megabytes a minute. Round that up and an hour of talking runs roughly three hundred and fifty to five hundred and fifty megabytes. The cheapest hour I recorded was closer to two hundred and fifty; the most expensive was a little over seven hundred, and I know exactly why, which I will come to. If you want one figure to carry around, carry half a gigabyte an hour, and treat everything else in this piece as a reason that figure moved.

Half a gigabyte an hour is worth setting next to things you already have a feel for. An hour of streaming music costs a small fraction of it. An hour of scrolling a vertical video feed — the thing most of us do without noticing we started — can easily cost more than the call did, because that feed is an unbroken run of fresh, high-bitrate clips with nothing repeated and nothing still. An hour of high-definition television streaming costs considerably more again — that one I have not measured myself, but nobody disputes it. Video calling has a reputation as the expensive thing on your phone, and on the evidence of my own counter that reputation is about one notch harsher than it deserves.

Turning cameras off changes the arithmetic completely rather than incrementally. Voice-only calls came in at roughly half a megabyte to a megabyte a minute, which is thirty to sixty megabytes an hour, which is to say that an entire evening of talking costs less than sending a couple of photographs. Text is barely worth measuring at all. A long typed conversation, the kind that starts late and ends later, hardly registers on the counter. If your allowance is genuinely tight this is not really a compromise; it is a different product that happens to live behind the same button.

The reason the video figure wanders so much is the way video compression works. The encoder is not sending a fresh photograph sixty times a second — it is sending what changed since the last one. A person sitting reasonably still in front of a plain wall changes very little from frame to frame, so the stream costs almost nothing to keep up to date. The same person waving their hands, walking between rooms, or wearing a shirt with a fine pattern that shimmers as they move, generates a great deal of change and pays for all of it. My seven-hundred-megabyte hour was a call I took while tidying the kitchen.

The finding that genuinely surprised me was about light. A dark room does not save data; it costs data. When there is not enough light the camera sensor compensates and produces noise, and that noise flickers randomly from one frame to the next. To an encoder, random flicker is indistinguishable from movement, so it dutifully spends bits describing it. A grainy, dim picture can therefore be simultaneously more expensive and worse-looking than a well-lit one. Aiming a lamp at your own face is the rare adjustment that improves the picture and reduces the bill in the same gesture.

The network gets a vote too, and it does not consult you before casting it. A modern call adapts to the connection it finds: on a weak signal the stream quietly steps down to something smaller and softer, and on a strong one it climbs back up. This is why two calls of identical length can differ by a factor of two with nothing else changed at either end. It also means a low reading is not automatically good news. Sometimes a small number is the record of a call that looked soft the whole way through, because the connection never gave it permission to climb.

Screen size plays a quieter part than people assume, though it does play one. A call filling a tablet asks for a larger picture than the same call in a phone-sized window, and the stream will happily supply it if the connection allows. It is not a dramatic difference — nothing like the gap between video and voice — but it is one reason the same conversation can cost more on the device you took it on at home than on the one in your pocket. If you are counting carefully, the small screen is the cheap screen.

Something worth stating plainly, because it is the lever almost nobody thinks to pull: you are paying in both directions. The video you send and the video you receive both cross your connection and both land on your bill. A one-to-one call is at least honest about this — there is one picture coming in and one going out, and nothing multiplies. Group rooms are where the arithmetic gets away from people, because every additional face is another stream arriving, and four faces on a screen is not four times the conversation but it is close to four times the bandwidth.

This is the point where I should correct something I assumed for most of that month, because it turns out to be wrong here and it is the kind of thing nobody tells you. Switching the camera off inside a video call on this site is a cover, not a tap that stops the stream. The control lives behind the more button on the call screen, in a small panel labelled Operate holding exactly two switches, microphone and camera; flipping the camera one hides the picture and tells the other side your camera is off, and the call remains a video call underneath. It changes what is seen. It does not, on its own, change what is sent.

The genuinely cheap option is a different thing entirely, and it is a separate way of starting: a voice call, placed from the private message screen rather than from the video button, builds only a microphone track. No camera is opened at all, the other person's avatar fills the screen instead of a picture, and the counter behaves accordingly. If your allowance is the binding constraint, that is the door to use — not the camera switch inside a video call, which is a privacy control wearing a data-saving costume. Look at who is online, open a message, and start from there.

If you go looking for other numbers you will find figures much larger than mine, and they are not wrong — they are measuring something else. The figures that circulate mostly come from desktop conferencing, where a call fills a large screen at high definition, frequently with several participants at once, on software tuned for a laptop sitting on home broadband. A one-to-one call on a phone, in a portrait window a few inches tall, is a different animal working at a fraction of the pixel count. Figures from the meeting-room world do not transfer into your pocket, and quoting them at each other is how the myth got started.

Turning all of this into something you can actually decide with: if you have a gigabyte of allowance left, you have roughly two hours of video, or fifteen-odd hours of voice, or a practically uncountable amount of text. Knowing that in advance is precisely what stops you doing what I did on that first call, which was to spend the good half of a conversation computing instead of listening. The arithmetic only needs doing once, and then it stops being arithmetic and becomes a habit you no longer notice.

A word about plans that call themselves unlimited, since a good share of people reading this are on one. In practice a great many of them are unlimited in the sense that you are never cut off and metered in the sense that, past a certain number of gigabytes, you are moved to the back of the queue whenever the local network is busy. You will not see a charge. You will see a call at eight in the evening quietly deciding it can only manage a soft picture. Knowing roughly where your own threshold sits explains a surprising amount of otherwise inexplicable call quality in the last week of a month.

Two traps ate more of my allowance than any actual call. The first is a phone quietly deciding your Wi-Fi is unreliable and switching to mobile data mid-conversation. Both major platforms will do this on your behalf, and the switch that controls it is buried somewhere most people have never opened. I lost around three hundred megabytes that way in a single evening at a friend's flat, two floors and several walls away from the router, and did not discover it until the monthly summary arrived and disagreed with my notes.

The second trap is everything else on the device. Photo backup deciding that now is the moment to upload a weekend away. An app store fetching updates because it saw a connection and got excited. Another device streaming through the same tethered hotspot. All of it lands in your monthly total, none of it is the call, and the call gets blamed anyway because the call is what you were doing at the time. Before concluding that video chat is ruinous, check what else woke up while you were talking. Usually something did.

The only number that really matters is the one your own device reports, and both platforms will tell you if you ask them properly. On an iPhone the cellular settings screen lists usage per app, with the important catch that the total never resets by itself — it counts from whenever you last reset it, which for most people is never, so reset it before you measure anything or the figure is meaningless. On Android the data usage screen is tied to your real billing cycle and breaks down per app, which quietly makes it the better instrument of the two. Either way, one call and five minutes gives you a figure for your phone, your plan and your habits, which is worth more than mine.

Two situations deserve more caution than the ordinary case. Roaming is the obvious one: the megabytes behave identically but the price does not, and a half-gigabyte hour that costs nothing at home can be genuinely expensive across a border, which makes it a Wi-Fi occasion by default. Connections on aircraft are the other. They tend to be metered severely and shaped to discourage video outright, which is why a call that connects at all up there usually arrives as audio with a slideshow attached to it.

My own habits settled into something unremarkable. Video at home, where the connection costs nothing and the light is decent. Voice first when I am out, with the camera arriving later if the conversation earns it. And I stopped thinking about the install side entirely, because there is nothing to download here in the first place — no hundred-megabyte package before you can say hello, and no background updates turning up on a mobile connection three weeks later when you are somewhere expensive.

There is one trade that these guides tend to leave out. You can push the data cost down by degrading the call — a smaller picture, a lower frame rate, worse light because you stopped caring — and the number will duly fall. But a stuttering, blocky picture is not a cheaper video call. It is a worse conversation with a video-shaped decoration attached. If it comes to that, an honest audio call at full quality beats a video call held together with string, and it costs a twentieth as much while sounding better.

The month finished with about forty calls in the notes app and, once I had subtracted everything that was not actually a call, a little under nine gigabytes across all of them. That is a figure I would have found alarming in advance and found completely unremarkable in hindsight, which is more or less the point of measuring anything at all. It also included two hours of one conversation that I would happily have paid for twice, and about forty minutes of a call that I would have liked to be billed for by the sentence.

What the month actually taught me had less to do with megabytes than with attention. The calls I remember are the ones where none of this crossed my mind — where the light happened to be fine, the connection happened to hold, and neither of us mentioned either. The calls I remember badly are the ones where I was watching a counter in the corner of my eye. If measuring it once buys you the right to stop thinking about it, that is the entire return on the exercise. The same is true of most of the small mechanical worries that sit between a person and a conversation, which is roughly why the other notes and guides here exist at all.

So: a few hundred megabytes an hour for video, a rounding error for voice, and most of the difference sitting with you rather than with the app. If you want to know where your own phone lands, reset the counter, start a call on voice, and place a video call afterwards to see the difference. It is free to start, and the first number you write down yourself will be more useful than any figure in this piece, mine very much included.