Recording

How to record multiple tracks in a browser

Yes, it genuinely works now: open a tab, record a part, record a second part over it while the first plays, and keep going. The browser stopped being the limitation years ago — AudioWorklet, the low-level audio processing a real recorder needs, has been available in every major browser since April 2021. What decides whether your recording sounds good is not the browser. It is three things: closed-back headphones, direct monitoring, and setting the gain before you play, every time.

By MELD · 12 September 2026 · 9 min read

What do you actually need?

Less than the gear lists suggest. For a first multitrack session:

• A laptop and a browser. Chrome, Edge and Firefox are all fine. Safari works but has real gaps — see the last section.
• Closed-back headphones. The one genuinely non-negotiable item, and the cheapest problem to solve. Why is the next section.
• An audio interface, if you are recording a microphone or an instrument. Any of them. The reason to own one is not fidelity, it is the headphone socket with direct monitoring on it.
• A microphone or an instrument cable. A dynamic microphone is far more forgiving in an untreated room than a condenser, which hears every reflection off your bare walls.

You can skip the interface and record from the laptop's built-in microphone. That is fine for capturing an idea at one in the morning. It does not work for overdubbing, because of the problem below.

Why closed-back headphones, specifically?

Because of a loop that quietly ruins everything after take two.

If you record a second part while the first plays through speakers, your microphone picks up the speakers as well as you. The first take is now baked into the second one. Mute the first track and it is still there, faintly, underneath. Every overdub after that adds another smeared copy, and by the sixth track the mix has a fog in it that no amount of EQ removes.

Closed-back headphones stop the loop dead. Open-back headphones leak enough to cause it. This is the single most common reason a home multitrack recording sounds muddy for no apparent reason.

How do you record the second track over the first?

The loop is the same in every recorder ever made, browser or not:

1. Set the tempo before you record anything. Guessing it afterwards and time-stretching to match is misery you can avoid with one decision at minute zero.
2. Record a scratch take of the whole song, all the way through, however rough. It is your map: it carries the arrangement, the real tempo feel and the section structure.
3. Arm a new track, check the gain, record the real part. Check the gain every single time. Singers move; a vocal level does not match the acoustic guitar level you set twenty minutes ago.
4. Listen back solo, then in context. Keep it or do it again. Redoing take one is cheap. Redoing it after six tracks were timed against it is not.
5. Repeat. Drums or a click first, then bass, then rhythm parts, then leads, then vocals last — the singer performs better against a filled-out arrangement.

Name the tracks as you go. "Audio 1, Audio 2, Audio 3" costs you real time by the third session and there is no way to recover it later.

Why does my overdub feel late?

Because it is, and the fix is not a setting — it is taking the computer out of the listening path entirely.

Sound goes from your instrument into the interface, gets converted, travels over USB, waits in the operating system's buffer, waits in the browser's, gets processed, and then makes the whole trip back to your headphones. Each stage costs milliseconds. Most people start noticing the round trip somewhere between 10 and 20 ms, and there is a useful physical intuition for it: sound travels about a foot per millisecond, so 20 ms of delay feels like your headphones are ten feet from your mouth. That is why overdubs recorded at high latency drag — you are unconsciously compensating for a delay.

Direct monitoring is the fix. Almost every interface has a switch that sends your input straight to its own headphone output, bypassing the computer. You hear yourself instantly, and the backing tracks come from the computer as normal. Turn it on, then turn input monitoring off in the recorder so you do not hear yourself twice — once instantly and once late, which produces a hollow comb-filtered sound that makes people think their interface is broken.

On the browser's side of it, the numbers are better than their reputation. An AudioWorklet processes audio in blocks of 128 frames — under 3 ms at 48 kHz — and has done since it shipped. The browser is not where your latency is coming from. Your buffer settings and your monitoring path are.

Can you record a whole band at once?

A recorder can hold any number of tracks. Recording several things simultaneously is limited by the physical inputs on your interface: two inputs means two things at once, whatever the software says.

For four inputs or fewer, a browser is a perfectly good tool. Past that, the honest answer is to use an installed recorder with proper multi-channel driver support — and a browser-specific reason, not just a general one: the constraint that lets a web page ask for more than one channel from a device, channelCount, is not supported in Safari at all, so multichannel capture is not portable.

There is a better approach for most bands anyway, and it is not a consolation prize — it is how a great many professional records get made. Play the song together once, recorded simply, to capture the feel and the tempo. Then overdub each instrument properly against that. Two inputs and patience beats eight inputs and an argument about headphone mixes.

Which browser should you use?

Chrome or Edge first, Firefox second, Safari last — and for reasons more specific than the usual hand-waving.

Recording needs two things beyond the basics: choosing which device to record from, and turning off the processing meant for video calls. A page asks for a specific interface with deviceId: {exact: id}, and asks the browser to leave the signal alone with echoCancellation: false, noiseSuppression: false and autoGainControl: false. Echo cancellation and automatic gain are catastrophic on music — they duck and pump a performance in ways that cannot be undone.

In Safari, noiseSuppression and autoGainControl are not supported constraints at all, and neither is channelCount. You can switch echo cancellation off and that is the extent of it. Output-device selection is worse: setSinkId is Chrome-only, so no page can choose which of your outputs it plays through in Safari or Firefox.

A recorder should always check navigator.mediaDevices.getSupportedConstraints() rather than assuming. A constraint a browser does not know is quietly ignored, which is how a take ends up with noise suppression chewing on it.

When is a browser the wrong tool?

It is worth being straight about this, because the answer is not "never".

• Eight or more simultaneous inputs. A full drum kit with the band isolated. Use an installed recorder and a proper interface.
• Playing sampled instruments live from a keyboard, where feel matters at the millisecond level. Programming MIDI with a mouse is fine; expressive live playing still feels better native.
• Very heavy sessions. A hundred tracks with large sample libraries will bring a tab down. Most songs never come near this; film scores do.
• A sound built on specific third-party plugins. Browser tools use their own effects. If your sound lives in a particular compressor you own, that is a native-recorder life, and that is legitimate.

For solo songwriting, overdubbing, sketching and finishing a song with a handful of parts — which is most music made by most people — the browser is no longer the compromise option.

Doing it in MELD

MELD Studio records from your microphone, lays takes onto a timeline, and exports a lossless WAV, in a tab, with nothing installed. Signed out you can record a take, trim it, fade it, normalise it and export it — no account, no card. The timeline, the mixer and the EQ you drag by its curve come with a paid plan.

Nothing you record is uploaded by recording it. A take lives in your browser's own storage until you press a button that says otherwise, and when you do want it heard, it goes from the timeline to MELD Master and out to a post in two presses.

Back your work up anyway, wherever you make it. Export a mix at the end of every session to your own drive. That habit costs ten seconds and it is the cheapest insurance in music.

Once the song is finished, the next question is how loud to make it: what loudness to master to, and which of the numbers you have read are actually published by anybody. And if the parts are coming from two people in two places, what decides whether you can play together online is distance more than software.