How to Calibrate Audio Offset in Rhythm Games Properly

When I moved flats last year I bought a new television, set it up, opened my usual rhythm game, and played the worst session I had played in months. My accuracy on a chart I knew well fell from 97.8 percent to 94.1. I blamed the move, then tiredness, then the game. It took two evenings to find the actual cause: the TV shipped with its game mode switched off, and its picture processing was adding enough delay that every note I hit landed late in the game's eyes.

Most rhythm game players run into a version of this eventually, often after changing headphones or screens. The game asks you to calibrate, you tap along to a metronome, it gives you a number, and somehow things still feel off. That is because there are two separate delays in play, and the standard tap test only measures one of them well, and also measures you.

Two delays, not one

Every rhythm game has to line up three things: the sound reaching your ears, the picture reaching your eyes, and your input reaching the game. Each travels a different path with its own delay.

Sound delay comes from your audio chain. Wired headphones add almost nothing. Ordinary Bluetooth headphones are the big offender; in my own tests and in most published measurements, standard Bluetooth sits somewhere between about 150 and 250 milliseconds, which is many times larger than any timing window in a serious rhythm game. Low latency codecs help, but only if both the headphones and the device support them.

Picture delay comes from your display. A monitor might add a few milliseconds. A TV outside its game mode can add 50 to 100 or more, because of the smoothing and sharpening it does to every frame. That was my problem.

Most games therefore offer two settings, usually called audio offset and visual offset, or audio and video calibration. Changing the wrong one makes the other problem worse, which is why "I calibrated and it still feels wrong" is such a common complaint in community forums.

Why tap tests mislead

The usual calibration screen plays a steady beat and asks you to tap along. It averages your taps and sets the offset to match. That sounds sensible, but it carries a quirk from how people keep time.

Researchers who study timing have known for decades that people tapping along to a metronome tend to tap slightly ahead of it, typically by a few tens of milliseconds. The effect even has a name in the literature, negative mean asynchrony. Musicians show less of it than non musicians, but almost nobody lands exactly on the beat. So a tap test measures your hardware delay plus your own habit of anticipating, and then bakes that habit into the offset.

That is not always bad. If you anticipate in the tap test, you probably anticipate in play too, and correcting for it helps. But it means the tap result changes with mood and fatigue. I have run the same test three times in an evening and got results 12 milliseconds apart.

Calibrate the hardware once, carefully. Then judge your timing from real play, not from a metronome screen.

The method I use now

This is the process I follow whenever I change a screen, headphones or console. It takes about twenty minutes and has not let me down since the TV incident.

1. Fix the hardware first. Turn on the TV's game mode, plug in wired headphones if you can, and switch off any audio enhancement on the device. Calibration cannot rescue a 200 millisecond Bluetooth delay in a game with a 30 millisecond window.
2. Set video before audio. Most games show a flashing marker for visual calibration. Get that matched first, since your eyes are more sensitive to drift on screen than your ears are to audio offset.
3. Use the tap test once, as a starting point. Accept its number, but treat it as rough.
4. Play three songs you know well, at a comfortable difficulty. Pick ones you can almost full combo, so mistakes come from timing, not from reading the chart.
5. Read the hit error display. Games like osu! show a bar of where each hit landed relative to perfect. If the cluster sits consistently to one side, shift the offset by a few milliseconds in that direction and play again.
6. Stop when the cluster centres. Spread is your skill. Centre is calibration. Do not try to fix spread with offset.

In osu! you can adjust the offset for a single beatmap before playing with the plus and minus keys, which is useful for the odd chart that was timed slightly off by its mapper. Change the global setting only when every chart feels off in the same direction.

How small the windows really are

It helps to know what you are aiming at. In osu!'s standard mode, the window for a perfect hit is calculated as 80 minus six times the chart's Overall Difficulty, in milliseconds either side of the note. At Overall Difficulty 8, that is plus or minus 32 milliseconds. A Bluetooth delay of 180 milliseconds is more than five of those windows. A TV without game mode can eat two or three of them.

That arithmetic is why calibration matters more in rhythm games than almost anywhere else. In an action game, 40 milliseconds of extra delay feels sluggish. In a rhythm game at a high difficulty, it is the difference between a perfect and a miss on every single note. For more on how the genre trains your ear and hands, see how rhythm games teach a sense of timing.

Hardware that makes calibration easier

I would not tell anyone to buy expensive gear for this. Cheap wired earbuds have less delay than expensive wireless headphones, and a basic monitor will usually beat a premium TV for timing. If you are building a setup on a tight budget, a budget rhythm gaming setup covers what is worth paying for. Some older band games went further and built light sensors into their guitar controllers so the game could measure screen delay by itself, which is still one of the cleverest solutions I have seen.

One more thing I learned from the TV: settings can reset after firmware updates. If your timing suddenly drifts for no reason, check game mode again before you doubt your hands.

A small task for this week

Pick one song you know very well and play it three times with the hit error display on, without changing anything. Note where the cluster sits each time. If it is off centre in the same direction all three times, move the offset five milliseconds that way and play once more. If it moves around between runs, your calibration is fine and the work left is practice, which is better news than it sounds. More on the music side of games lives in our games section.