Home Vocal Recording and Mixing

Why your vocal sounds thin and quiet after uploading to YouTube, Instagram or TikTok

By StudioRescue Editorial

Published September 28, 2026 · Updated September 28, 2026

This is one of the more demoralizing moments in home recording. The vocal sat right in the mix, the export sounded good on your own system, and then the uploaded version comes back smaller, duller, and further away. The instinct is to go back and add brightness or more compression, which usually makes the next upload worse.

Before changing anything, it is worth separating three different things that all get described as “it sounds worse after uploading”: a level change, a processing decision you already made, and distortion created during the platform’s re-encode. They have different fixes, and only one of them is the platform’s doing.

First, prove what actually changed

The whole diagnosis rests on one question: did the sound change, or did the level change? An honest comparison answers it in a few minutes.

Why quieter is heard as thinner

This part is physiology, not audio processing. Human hearing is not equally sensitive across the frequency range, and the shape of that sensitivity changes with level. As playback gets quieter, the low end and the extreme top drop away faster than the midrange does. The international standard that describes this, ISO 226, publishes the equal-loudness contours that map it out.

The practical consequence is that a reduction of only a few dB does not read to your ear as “the same thing, slightly quieter”. It reads as smaller, thinner, and less exciting. So a platform that turns a hot upload down by 3–5 dB will make a perfectly good vocal sound weaker, even though every frequency is still exactly where you left it.

That is why the level-matched comparison in step 3 matters more than any plugin. It is the only way to find out whether you have a tone problem at all.

Why “crushed” is a different problem, and your own

If the upload sounds flat, lifeless, or pumping rather than simply quieter, look at what you did to get it loud. Heavy limiting flattens the peaks of a vocal, reduces the difference between the loud and quiet parts of a performance, and can add a squashed, breathing quality.

Here is the part that catches people out. The platform turns your loudness down, but it does not undo the limiting. You keep every bit of the damage and lose the loudness it bought you. Independent write-ups of YouTube’s behavior also agree on a related point: it reduces uploads that are above its reference and leaves quieter ones alone rather than raising them. Pushing harder than the reference is therefore a one-way trade that only costs you.

If your vocal is inconsistent in level, the answer is gentler, earlier control rather than a heavier limiter at the end. Compression on the vocal track, and automation where compression is too blunt, both do that job without flattening the master.

The third cause: peaks that only appear after encoding

Uploads get re-encoded into a lossy format for delivery. When a lossy file is decoded for playback, the reconstructed waveform can rise slightly above the highest sample values in your original file. These are inter-sample peaks, and they are what true-peak metering, in dBTP, is designed to catch.

If you rendered with peaks sitting at or very near 0 dBFS, those inter-sample peaks can clip on playback. That shows up as grit or harshness on the loudest words and on sibilance, and it is genuinely new distortion rather than a level illusion. Leaving headroom prevents it.

Two published references are useful here. The EBU R 128 broadcast recommendation sets a maximum permitted level of -1 dBTP, measured to ITU-R BS.1770. Spotify, which publishes its own guidance, recommends a true peak below -1 dBTP for lossy delivery, and below -2 dBTP for masters louder than its -14 LUFS reference. Neither is a YouTube or TikTok specification, but both point the same way: a little headroom costs nothing and prevents encode distortion.

Numbers that are published, and numbers that are not

Search results on this topic are full of confident tables of “official” loudness targets for every social platform. Most of them are worth very little, so it is worth being clear about what is actually documented.

Independent estimates of YouTube’s reference level exist and are commonly quoted somewhere between -16 and -13 LUFS, but they disagree with each other and none of them comes from YouTube. That is precisely why the Stats for nerds reading is more useful than any target: it is the platform’s own measurement of your file, not somebody’s estimate of the platform.

For Instagram and TikTok there is no equivalent readout. The honest method there is a level-matched listen, plus keeping your true peak under control so the encode has room to work.

What to change in REAPER before the next upload

1. Measure before you render. Use Dry Run in the Render dialog and read the integrated LUFS and true peak. Guessing is how this problem starts. The full walkthrough is in how to export audio from REAPER.
2. Stop aiming for maximum. Anything above the platform’s reference gets taken back off. Deliver a clear, consistent vocal at a sane loudness instead. For a voice-led video, the target ranges and setup are covered in how to record a voiceover in REAPER for YouTube. Podcast delivery uses different targets again, so do not carry one over to the other.
3. Leave peak headroom. A true peak at or below -1 dBTP is a safe starting point for anything that will be re-encoded.
4. Back the limiter off and re-check. Render again, compare at matched level against the crushed version, and keep whichever one sounds better when neither is louder.

What this will not fix

• It will not make a thin recording sound full. If the vocal was already lacking body before it left REAPER, that is a tone problem at the source, covered in how to fix thin vocals in REAPER.
• It will not make you louder than other creators. That is the point of normalization. Everyone above the reference is pulled back to it.
• It will not repair a source that was already clipped on the way in. See fixing clipped and distorted audio in REAPER.
• It will not change how a phone speaker reproduces low frequencies. A small speaker will always sound thinner than your headphones, whatever you upload.

If it still sounds wrong

1. You did not match levels. Repeat the comparison properly. This resolves most cases on its own.
2. Your video editor re-encoded it. If you exported audio from REAPER and then rendered a video, the editor has its own audio settings and may have resampled or re-compressed the track. Check the file you handed it against the file the editor produced.
3. The upload is playing in mono. A stereo file folded to mono can lose level and body through phase cancellation. If one side sounds missing, see fixing audio recorded in mono or on one channel.
4. It is harsh rather than thin. Harshness that appears only after upload points at the encode, so lower your true peak and render again before touching EQ.