AUDIO QC BLOG / 8 MIN READ

How do I check my audio levels before uploading?

Bad audio loses more viewers than a soft shot ever will, and you cannot trust your own ears to catch it. Here are the two numbers that matter, the meter to read them on, and three ways to run the check.

−14 LUFSYouTube target
−1 dBTPpeak ceiling
2numbers to read
0 to 100craft score

By Thomas, founder of CutScore · Updated June 2026

AUDIO CHECK · final_mix.mp4
Two people leaning over a lit audio mixing board, reading the level meters on a finished mix before it goes out, the exact moment an audio check should happen.
CRAFT SCORE
FIXES ADVISED
read the meter, not the room
Too quiet · −19 LUFS, lift +5 dBwhole
True peak safe · −1.4 dBTP
Music over voice · duck −4 dB00:42
The 30-second answer To check your audio levels before uploading, put a loudness meter on your finished export and read two numbers. Integrated loudness should land near −14 LUFS for YouTube (a little quieter, around −16 to −14, for vertical feeds), and true peak should stay at or below −1 dBTP so nothing crackles after the platform re-encodes your file. Then do one last listen on the cheapest speaker you own and confirm the voice stays clearly on top of the music. If checking those by hand sounds like a faff, that is the exact job CutScore does in one pass.
WHY YOUR EARS LIE TO YOU

I have shipped videos that sounded great in my headphones and tinny everywhere else. The problem is not your ears, exactly. It is that you heard the mix a hundred times in the edit, and your brain stopped flagging the hum and the harsh "s" sounds hours ago. By export time you are not hearing the audio. You are remembering what you meant it to sound like.

Then your gear flatters you. Good headphones add bass and clarity that a three-year-old phone, one tinny speaker, on a noisy train, simply does not have. So the mix that felt warm and full on your desk arrives thin and quiet for half your audience. Loudness is the worst offender, because there is no visual cue for it. A dark shot looks dark. A quiet mix just sounds normal, until it plays right after a video that is twice as loud.

So the answer is not "listen harder." It is measure, then listen on bad speakers. Listening is a vibe. Measuring has targets. The two targets that matter are short, dull, and almost entirely under your control. Here they are.

THE TWO NUMBERS

Loudness and true peak, in plain terms.

Forget the dozen meters in your editor. For a pre-upload check, two readings carry almost all the weight. Hit both and your audio sits at the right level without distorting.

ReadingTarget to hitWhat it costs you if you ignore it
Integrated loudness (YouTube)≈ −14 LUFSToo quiet and the video feels weak and timid next to everything else in the feed.
Integrated loudness (Reels, TikTok, Shorts)≈ −16 to −14 LUFSVertical feeds normalise a touch quieter; aim slightly lower than YouTube so it is not blasting.
True peak≤ −1 dBTPHot peaks crackle and distort once the platform re-encodes your file to its own codec.
Voice vs musicvoice clearly on topMusic burying the speech is the single most common amateur tell in the whole feed.
Noise floorlow, steadyConstant hiss or hum reads as "cheap" before a single word is understood.
One thing the platform does for youYouTube and the rest normalise loudness on playback, so they will turn a too-loud video down. They will not turn a too-quiet one up. That asymmetry is why aiming for −14 LUFS matters: undershoot and you stay quiet forever; overshoot and you only lose dynamic range.
SKIP THE MANUAL READ

Reading a loudness meter on every export gets old fast. CutScore measures integrated loudness, true peak and the voice-to-music balance in one pass, then tells you the exact gain change to make.

Join the waitlist
HOW TO ACTUALLY READ THE LEVELS

Four quick passes, in order.

1. Put a loudness meter on the finished export

Check the file you are about to upload, not a clip mid-edit, because the integrated reading is measured across the whole thing. Any EBU R128 or ITU-R BS.1770 meter works. Premiere and DaVinci Resolve both have one built in, Audacity ships a loudness analyser, and the free Youlean Loudness Meter shows integrated LUFS and true peak side by side. Play the whole video through it. The number you care about is integrated (sometimes labelled "program" or "I"), not the momentary value bouncing around. Aim for −14 LUFS on YouTube, a touch quieter on vertical feeds.

Hands resting on the faders of a large audio console under studio light, the kind of metering view where integrated loudness and true peak get read before a mix is signed off.
Read the integrated value across the whole file, not the meter bouncing on a single line. Photo: Tima Miroshnichenko / Pexels.

2. Check the true peak, not just the loudness

Loudness and peak are two different things, and people confuse them constantly. A mix can read a polite −14 LUFS on average and still have a single transient (a clap, a hard consonant, a door slam) that spikes right up to the ceiling. When the platform re-encodes your file, those spikes can push over 0 and crackle. Keep your true peak at or below −1 dBTP. If your meter shows a true peak above that, a limiter with the ceiling set to −1 dBTP fixes it without changing how loud the video feels. If you are not sure the spikes are distortion, here is how to tell whether your audio is clipping.

3. Balance the voice against the music

Numbers do not catch this one, your worst speakers do. The most common audio fault I see is not a wrong LUFS value at all. It is music sitting on top of the speech, because the creator mixed it on headphones where every word was crystal clear. Solo the music, then bring it back, and ask whether you would still understand every sentence on a phone speaker. If the answer is no, pull the music down four or five decibels and duck it further under the talking. Better yet, automate it: louder in the gaps, quieter under the voice. If yours already drowns the talking, this is why your music is louder than your voice.

4. Do the bad-speaker listen

This is the step that catches what the meter cannot. Export, then play the video back on the cheapest, nastiest speaker you have: a laptop, a phone held at arm's length, one earbud on a train if you can. Listen for three things. Can you hear every word over the music? Is there a hum or hiss you tuned out hours ago? Does the whole thing feel about as loud as the last video you watched on that device? If yes, yes (the bad kind), and quieter, you have your fixes. For the full picture on why quiet audio happens in the first place, see why your video audio is so quiet.

RATHER SEE IT THAN READ IT?

Here is a real CutScore report for an everyday talking-head video: loudness, true peak, voice-to-music balance and the rest, scored, with timestamps and the exact fixes.

See a sample report
SHORT ON TIME

If you only fix three things.

Most of the jump from "sounds homemade" to "sounds professional" comes from these three. Do them in this order.

1
2-MIN FIXAUDIO
Normalise the mix to about −14 LUFS
Quiet audio is the fastest way to sound amateur, and it has nothing to do with your microphone. Normalise the whole export toward −14 LUFS for YouTube (a touch lower for vertical) and your video instantly sits at the same level as everything around it.
How Run a loudness meter over the export and apply the gain it suggests, or let CutScore measure it and tell you the exact change.
2
QUICKAUDIO
Cap true peak at −1 dBTP
Pop a limiter on the master bus with its ceiling set to −1 dBTP. This stops hard consonants and transients from clipping after the platform re-encodes your file, and it does it without making the video feel quieter.
How Add a true-peak limiter as the last thing in the chain. See true peak.
3
MIXAUDIO
Get the voice clearly above the music
Half your viewers decide in seconds, and they decide on the words. If the music competes with the speech, pull it down four or five decibels and duck it further under the talking. The voice should win on a phone speaker, not just in your headphones.
How Solo the dialogue, set its level, then bring music back underneath it. Listen on your worst speaker.
THREE WAYS TO RUN THE CHECK

By ear, by meter, or in one pass.

OPTION 01

By ear alone

Free, and the least reliable. Your ears adapt and your gear flatters, so a mix that sounds full on your desk can arrive thin on a phone. Works best on someone else's video, or yours after a day away. Always finish with a listen on your worst speaker.

OPTION 02

With a loudness meter

Accurate and honest. A loudness meter on the export gives you integrated LUFS and true peak in two numbers. The cost is time and knowing the targets: you have to open the tool, read it right, and do it for every video. Great if you enjoy this. Most people do not.

OPTION 03

With a coach in one pass

Hand the file (or a link) to CutScore. It measures loudness, true peak and the voice-to-music balance against the right standard for your platform, then gives you a 0 to 100 score with timestamped evidence and the gain changes. No meter to read. See a sample report.

How CutScore checks your audio for you CutScore is an AI video quality coach for pre-publish QC. It measures loudness with an EBU R128 meter, reads true peak in dBTP, and estimates the balance between voice and music, all deterministically, so the numbers are repeatable rather than a guess. You get one score, the evidence behind it, and the exact gain change to hit your target, before anyone else hears the video. It judges the craft of the audio itself, so it sits happily next to a growth tool rather than competing with one. More on the method and the standards.
QUESTIONS

Frequently asked.

Put a loudness meter on your finished export and read two numbers. Integrated loudness should sit near −14 LUFS for YouTube, and true peak should stay at or below −1 dBTP so nothing distorts after the platform re-encodes your file. Then listen on the worst speakers you own and make sure the voice stays clearly on top of any music. If all three hold, your audio is ready.
Any EBU R128 or ITU-R BS.1770 loudness meter works. Premiere and DaVinci Resolve have one built in, Audacity ships a loudness analyser, and free plug-ins like Youlean Loudness Meter read both integrated LUFS and true peak. The peak meter on your camera or phone is not enough, because it shows momentary peaks, not the integrated loudness a platform actually measures.
Your ears adapt and your gear flatters. You heard the mix a hundred times in the edit, and your laptop or headphones probably add bass and clarity that a phone speaker on a noisy bus will not. That is why you measure with a meter instead of trusting the room, and why you do a final listen on the cheapest speaker you can find.
Yes. CutScore measures integrated loudness, true peak and the balance between voice and music on your exported file, then reports the exact gain change to hit −14 LUFS with peaks under −1 dBTP. You get the numbers and the fixes with timestamps, before you publish, without opening a meter yourself.
EARLY ACCESS

Stop guessing at your levels.

CutScore reads your loudness and true peak for you and tells you exactly how much to lift or duck, with the evidence to back it up. Join the waitlist for early access.

Join the waitlist