Level Livestream Audio in the Browser: No Second PC
To level a multi-guest livestream in the browser, open BeamTracer's Loudness tab, point it at your microphone or the stream studio's tab audio, and set a speech target. Measure one guest at a time, adjust each one's gain until their integrated loudness sits on the target, and press RESET between people. No second computer, no loopback app.
Hi, I'm Trina with BeamTracer.com — here's how I'd level a multi-guest stream without buying a second computer.
There's a well-known tutorial workflow for getting a multi-guest livestream's voices to sit at the same level, and it goes like this: set up a second computer, route the stream studio's output into it with a paid loopback app, run a desktop loudness meter on that machine, and have each guest talk while you read their numbers off the second screen.
It works. People genuinely do this. And every piece of it exists because the meter and the stream couldn't live in the same machine without the audio routing turning into a science project.
A browser meter deletes the routing. The stream studio is a tab; BeamTracer's Loudness tab can listen to a tab; that's the whole rig.
Do you need a second computer to meter a livestream?
No. The second computer in the classic workflow exists only to host the meter, and a browser meter runs on the machine that is already streaming. Everything below replaces the second machine, the paid loopback app and the desktop meter with one tab.
| The two-computer workflow | One browser tab | |
|---|---|---|
| Hardware | A second computer | None — same machine that's streaming |
| Audio routing | Loopback/capture app (paid) | Tab audio or mic, picked in Settings → Audio |
| Meter | Desktop plugin or standalone app | The Loudness tab, free |
| Setup time | An afternoon, once | Under a minute |
| Reset between speakers | On the meter app | One button — same meter |
To be fair to the original: if you're producing broadcast with a hardware console and dedicated monitoring, a second machine has its place. But if your show is a browser-based studio and three guests on webcams, the second computer was only ever there to host the meter.
Why is LUFS the right measurement for matching voices?
Because LUFS averages the way your ears do and throws the silences out of the average. Peak meters are nearly useless for matching talkers — two voices can peak identically and sit 6 dB apart in perceived loudness. LUFS is gated, so the gaps between sentences are excluded and a hesitant guest and a fast talker are measured on their speech, not their pauses.
That gives each guest a single honest number — their integrated loudness — and matching guests becomes matching numbers.
Where −23 LUFS comes from
The tutorials level speech to −23 LUFS because that's EBU R128, the European broadcast standard's programme target — a genuinely good speech reference with decades of thinking behind it. But know what it is: a broadcast convention, not a livestream requirement. I couldn't find a published live loudness target from Twitch or any other consumer platform — as far as I can tell, live streams aren't normalized the way uploads are. If your stream mostly gets watched as a YouTube VOD afterwards, leveling nearer −14 is a defensible choice. What actually matters is that everyone on the show sits at the same number — pick your reference and be consistent.
How do you level multiple guests to the same loudness?
Set a speech target, switch the readings to LU so the target reads as zero, then measure one guest at a time and trim their gain by whatever their integrated number says. Everything here is free and runs in the tab: open the Loudness meter, and in Settings → Audio point the input at your microphone or at tab audio captured from the stream studio's tab — whichever place your mixed voices actually are.
- Set the TARGET card to EBU R128 (−23 LUFS) — or Custom if your show has its own house number
- Switch the readings to LU so every number is relative to the target and "correct" reads as 0
- Have one guest talk at their natural level for about 30 seconds, everyone else muted
- Read their integrated number, adjust their gain by that many dB in your studio, and confirm it settles near 0 LU
- Press RESET and repeat for the next guest — one clean measurement per person
- During the show, watch the ST UNDER counter: a climbing count means someone has drifted quiet for sustained stretches, not just paused
- Export the show's measurement as CSV or JSON when you're done, or use COPY for a numeric summary to paste into your notes
The LU relative mode is the underrated tool here. In absolute mode you're holding "−23" in your head and subtracting; in LU mode the meter does it, and every guest's reading is simply how far off they are. Guest reads −4 LU? They need 4 dB. Reads +2? Pull them down 2. The graph's target line becomes zero too, so the trace itself tells the story at a glance.
The RADAR view during a show: short-term loudness paints the disc over a one-minute sweep, the outer ring is a circular momentary meter, and the PEAK lamp in the corner remembers any true-peak over even if you looked away.
Can you trust a browser meter on live microphone input?
Yes, and for two specific reasons. Browsers love to "help" live audio — echo cancellation, auto-gain, noise suppression — and all of that is disabled on BeamTracer's analysis path, so the number describes your guest's actual level. The second reason is that interruptions can be excluded from the measurement rather than averaged into it.
BeamTracer measures the raw signal. The meter reads the signal before any of the app's own processing, so a guest's reading is not a processed version that won't exist on the stream.
Pauses are honest. Producer talkback, an ad break, someone's dog — press PAUSE and that stretch is excluded from the measurement entirely, rendered grey in the history so you can see what was left out. RESUME continues cleanly. Your leveling numbers never contain audio that wasn't the show.
The counters are your during-show alarm system. ST UNDER and ST OVER only count excursions sustained for at least five seconds, so a dramatic pause or one loud laugh never trips them — and silence never counts as "under". If a counter is climbing, it's telling you about a real drift worth a gain ride, which is exactly the thing that's hard to hear from inside a show you're also hosting.
What loudness should a podcast be, and does the recipe change?
The recipe is identical; only the number changes. The podcast world's convention is −16 LUFS for stereo programmes — it's the published requirement for Apple Podcasts — and −19 LUFS for mono. Set that in the TARGET card instead of EBU R128 and run the same steps.
Those aren't two different opinions, by the way. A mono file plays from two speakers, so it measures about 3 dB hotter than the "same" audio in stereo — −19 mono and −16 stereo are the same perceived loudness written down two ways.
Level at the source, fix nothing in the edit
Twenty minutes of soundcheck leveling saves the classic post-production afternoon of riding every guest's fader. If the voices went down matched, the edit starts finished — and if you do need proof of levels for a client, the meter exports the whole session as CSV or JSON, free.
Next
- LUFS in plain language — what integrated, momentary and short-term actually mean
- What Spotify actually does to your track — the same meter pointed at your masters
- How dynamic is your master? PSR, PLR and LRA — reading dynamics, not just level
- Best audio visualizers in 2026 — what else lives in this app
Level your next show with one browser tab. Free, nothing to install.
OPEN THE LOUDNESS METER →If this saved you a second computer, the meter's a tab away. beamtracer.com — see you in the next one.