Karaoke used to need a machine, a pile of discs and a living room big enough to squeeze everyone in. The remote version needs none of that. If five friends in five cities can each open a tab and hear one another, you already have a karaoke night. The rest is technique, and most of it comes down to three things: headphones, one person owning the music, and accepting that two people cannot physically sing at the same time over the internet.
No apps, no installers — just a browser, decent headphones and some discipline about the mic.
The minimum setup, honestly
- A browser on a laptop or a phone that can reach your microphone.
- A room where everyone can talk. You can start a voice room and share the link with your friends in about ten seconds. It is free, there is nothing to download, it runs in the browser, and there is no sign-up to slow anyone down — which matters when one friend is joining from a phone on a train.
- Headphones. Not optional. This is the single rule that decides whether your evening works or turns into a howling mess.
- A source for the backing tracks, held by one person at a time.
If someone’s microphone refuses to cooperate, the usual causes are a browser permission denied once and never asked again, or the wrong input device selected at system level; our guide to fixing a microphone that will not work in the browser walks through them in order. Sort that out before anyone sings, not during the second chorus.
Why headphones are genuinely not optional
Open speakers create a loop. Your speakers play the backing track and the other singers’ voices; your microphone picks that sound back up; your browser sends it out again; it comes back. That loop produces the rising howl everyone recognises, and even when it stops short of howling it leaves a smeared, doubled echo that makes the room unlistenable.
Browsers do ship acoustic echo cancellation, and it is good — at speech. It assumes one person talking into a quiet room, and works by modelling what left your speakers and subtracting it from what your mic hears. Music breaks that model, and the failure is ugly: the backing track pumps, drops out on every loud syllable, goes choppy in the chorus. Half the complaints about “bad music quality” are echo cancellation fighting the track, not a weak connection.
So: headphones on everyone, no exceptions. One person on speakers spoils the mix for the whole room, not just for themselves. Prefer wired ones. Wireless adds 50 to 200 ms of its own delay, and many wireless headsets drop to a low-bandwidth mono voice profile the moment their microphone activates — the same reason music suddenly sounds muffled on a call. Cheap wired earbuds beat expensive wireless ones here.
Monitoring: hearing yourself without hearing yourself twice
Sealed headphones create a second problem. With both ears covered you hear your own voice only through your skull. Bone conduction is bass-heavy and misleading, so singers drift off pitch and sing far louder than they think.
The studio fix is to leave one ear cup off, or half on. One ear takes the backing track, the other takes the room and your actual voice. It looks silly. It works, and it is the single change that most improves how people sing.
Do not solve this by routing your microphone back into your own headphones through the connection. Any monitoring path that goes out and comes back carries the full round trip, and a copy of your own voice arriving 100 ms late is classic delayed auditory feedback — it makes fluent speech, let alone singing, physically difficult. Monitor acoustically with an open ear, never electronically.
Who plays the music, and where it should live
This decision shapes everything else, and most groups get it wrong first time. The instinct is to have a host play every track for everyone. It works, but it creates a mismatch: listeners receive the music over one network hop and the singer’s voice over two, so the vocal lands consistently behind the beat — noticeably, unpleasantly late.
The stronger approach is that the singer plays their own backing track and mixes it into their outgoing audio. When music and voice are captured at the same point and travel as one stream, they arrive in perfect sync for everyone. There is still a delay, but it is a uniform delay of the whole performance, which nobody perceives as a fault. On Windows this is exactly what the loopback input exists for, and it is covered step by step in our guide to playing music in a chat room with Stereo Mix. Read it before your first session, not during it.
Two settings matter once music travels down the microphone channel. If your system exposes them, turn off noise suppression and automatic gain control on that input. Noise suppression treats sustained musical content as noise and eats it. Automatic gain control rides the level constantly, so the moment you go from a quiet verse to a belted chorus it clamps down, then pumps back up during the next quiet bar.
Latency: why two people cannot sing together
Mouth-to-ear delay on a normal connection runs roughly 80 to 250 ms once you count the capture buffer, encoding, the jitter buffer that smooths out network variation, decoding and playback. Musicians playing together need to stay under about 25 to 30 ms — past that, delay stops feeling like room ambience and starts feeling like a mistake. You are five to ten times over budget, and no setting fixes it, because much of it is the speed of light through fibre plus the buffers that keep audio from stuttering.
So design around it:
- One voice at a time. One singer live, everyone else muted.
- Relay duets. Trade lines instead of harmonising: one voice on the verse, another on the chorus. It sounds deliberate and sidesteps the problem entirely.
- The messy group chorus. Everyone unmutes for the last chorus and accepts the smear. Once a night, funny. Every song, exhausting.
- Applaud in text. Cheering over a live vocal just buries it. Type instead — the audio channel stays clean and the praise scrolls past nicely.
Managing the microphone queue
Remote turn-taking needs more structure than a real room does: you lose the body language that says someone is about to start.
Appoint a host who announces who is up and who is on deck, and nothing else. Post the running order in the text channel so nobody has to ask. The on-deck singer cues their track during the previous song, so there is no ninety-second silence while someone hunts for a file. Everyone uses hard mute, not “sitting quietly”, with thirty seconds of open mics for applause after each song.
Groups new to talking together warm up faster with a round of online icebreaker games first — nobody wants to be the one singing into a silent room. If you host regularly, set up a dedicated space instead of improvising each week; our walkthrough on how to create a chat room covers the settings worth fixing in advance. And if you want an audience beyond your own circle, voice chat rooms where you can talk to strangers give a karaoke night a very different energy.
Songs that work for a remote group
What works: mid-tempo songs with a chorus everybody already knows. Narrow range, ideally under an octave and a half, since you cannot change key on the fly. Call-and-response structures, which turn latency into a feature. Songs with instrumental breaks, which give the room space to talk. Anything where enthusiasm counts for more than accuracy.
What does not: fast rapped verses, where losing the beat by a fraction of a second is fatal. Tight harmonies — see the physics above. Songs built around one enormous held note, which is unkind to the singer and to the microphone. Obscure picks nobody can join in on, however good they are.
On keys: the track is fixed, so pick songs that sit in your range rather than songs you love but cannot reach. A confident performance three steps low wins over a strained one at pitch.
Volume: leading the mix without drowning it
Start with the backing track quieter than feels natural in your own headphones. A loud track in your ears makes you sing harder to compete, and a clipped microphone is ruined at the source — no listener’s volume control can undo distortion baked in before the signal left your machine. If you are mixing the track into your outgoing stream, keep it clearly below your voice: the vocal leads, the music supports.
Position matters as much as levels: keep the microphone about a hand-span away and slightly off to the side rather than straight in front of your mouth; that alone removes most of the popping on p and b sounds. On loud passages, pull back a few inches. And agree in advance that the host can say “you’re too loud” without it being rude: the person running hot is always the last to know.
A running order that works
Fifteen minutes of setup before anyone sings: everyone joins, checks their microphone, confirms headphones, one short level test. Then two hours in rotation, one song each, nobody taking a second turn until everyone has had a first. A group chorus to close, and text chat open throughout for requests, lyrics and applause.
The technology is the easy half. Respect the headphone rule, keep the music sitting with the voice that made it, and the only hard question left is what to sing.