An ordinary speed test says nothing about video calling. We simulate a real call: data up and down simultaneously at realistic bitrates, while measuring live whether your connection holds up.
A video call is technically the heaviest thing your connection does daily. Unlike streaming or downloading, your line must continuously send data in both directions simultaneously, without interruptions, with a delay of at most a few dozen milliseconds. A 500 Mbps connection with a latency spike every few seconds delivers a worse call than a stable 25 Mbps line.
This test therefore simulates a real call: we send roughly 3.5 Mbps up and down simultaneously — comparable to an HD group call — while measuring latency, jitter and packet loss every half second. That is exactly the traffic pattern video call apps generate, and exactly what an ordinary speed test never measures.
These are the official bandwidth requirements published by the vendors themselves. They are almost always lower than people expect.
| Platform | Scenario | Up | Down |
|---|---|---|---|
| Microsoft Teams | One-to-one video, minimum | 150 kbps | 150 kbps |
| Microsoft Teams | One-to-one video, recommended | 1,5 Mbps | 1,5 Mbps |
| Microsoft Teams | Meeting, recommended | 2,5 Mbps | 4,0 Mbps |
| Microsoft Teams | Screen sharing, recommended | 1,5 Mbps | 1,5 Mbps |
| Microsoft Teams | Audio only, recommended | 58 kbps | 58 kbps |
| Google Meet | 720p video | up to 1.7 Mbps | up to 1.7 Mbps |
| Google Meet | 1080p video | up to 3.6 Mbps | up to 3.6 Mbps |
| Google Meet | Group meeting | from 250 kbps | up to 4.0 Mbps |
| Google Meet | Audio only | 100 kbps | 100 kbps |
Source: Microsoft Learn, Prepare your organization's network for Teams, and Google Workspace Admin Help, Prepare your network for Meet meetings & live streams. Retrieved August 2026. We reproduce these vendor figures and do not independently verify them. Zoom does not publish its requirements in a form we can reliably cite; the order of magnitude is comparable.
Even the heaviest row asks for 4 Mbps. Virtually any modern connection clears that easily. That is exactly why “my internet is 500 Mbps” says nothing about whether you can video call well: the bottleneck is never bandwidth, it is what happens to your latency while those two streams run at once. This test deliberately measures at 3.5 Mbps per direction, well above the recommended figures, and watches the latency.
MOS stands for Mean Opinion Score: a 1 to 5 scale expressing expected call experience. Originally the score came from having people rate calls. We instead compute it from our measurements using the E-model from ITU-T Recommendation G.107.
Concretely: we estimate one-way delay as half your measured latency plus twice the jitter, convert that into a delay penalty, add a loss penalty of 2.5 points per percent of packet loss, subtract the total from a base value of 93.2, and convert the resulting R-factor into a MOS. That entire chain is in this page's source; you can verify it.
| MOS | Meaning | What you notice |
|---|---|---|
| 4,2 – 5,0 | Crystal clear | Even long group calls with screen sharing stay smooth. |
| 3,8 – 4,2 | Excellent | Smooth; small variations are not noticeable in practice. |
| 3,2 – 3,8 | Good | Fine, though heavy group calls may reduce video quality. |
| 2,6 – 3,2 | Fair | Occasional stutters, frozen video or robotic audio. |
| 1,0 – 2,6 | Poor | Structurally unusable for serious calls. |
The bands above are the conventional interpretation ranges as we apply them; G.107 does not prescribe exact labels.
Far less than most people think. Microsoft lists 1.5 Mbps in both directions as the recommended figure for a one-to-one Teams video call, with 150 kbps as the absolute minimum; a Teams meeting recommends 2.5 Mbps up and 4.0 Mbps down. Google lists up to 1.7 Mbps for 720p and up to 3.6 Mbps for 1080p on Meet. Audio only works on 58 to 100 kbps. In short: 5 Mbps in both directions is ample for virtually any call.
Because bandwidth is almost never the cause. A video call continuously sends and receives small packets, so it is mainly sensitive to latency, jitter and packet loss. The moment someone else in the house starts downloading, the queues in your router fill up and your latency spikes — bufferbloat. Your throughput stays high while your call becomes unusable. This test measures exactly that: what happens to your latency while traffic runs in both directions.
MOS is a 1 to 5 scale for expected call experience. From 4.2 it is crystal clear, 3.8 to 4.2 is excellent, 3.2 to 3.8 is good but may downshift on heavy group calls, 2.6 to 3.2 gives occasional stutters or robotic audio, and below 2.6 it is structurally unusable. We compute the score using the E-model from ITU-T Recommendation G.107 based on your measured latency, jitter and packet loss.
Jitter is the variation in packet arrival times. A steady 60 ms ping is fine for a call, because the software can absorb that delay. A ping bouncing between 20 and 200 ms is not: the receiver cannot size its buffer and discards packets that arrive too late. You hear that as robotic audio and see it as frozen video. Under 10 ms jitter is good; above 30 ms it becomes noticeable.
A cable, by a wide margin. WiFi causes the majority of all stuttering calls: it shares one medium with every other device and with your neighbours' networks, which directly produces jitter and packet loss. To find out whether a problem is your ISP or your WiFi, run this test once over WiFi and once over a cable. If the results differ sharply, your WiFi is the cause, not your connection.
Do not just measure speed, measure behaviour under load. This test does that in two phases: first five seconds of simultaneous download and upload at full speed to establish your capacity, then thirteen seconds of a simulated call at roughly 3.5 Mbps per direction with latency sampled every half second. Those measurements produce a MOS score and a verdict per scenario — one-to-one, group call, screen sharing.
Indoors it is usually WiFi interference or an overfull router queue; outside it can be a congested neighbourhood link or a problem at your ISP. Every lost packet is audio or video that never arrives: you hear words drop out or see the picture freeze. Above roughly 1 percent loss it becomes noticeable, above 3 percent calling becomes unpleasant. If you see loss over a cable while doing nothing else, measure at several times of day and take it to your ISP.