Home Vocal Recording and Mixing

Can this vocal take be fixed in the mix, or do you need to re-record it?

By StudioRescue Editorial

Published October 20, 2026

This is a triage, not a mixing tutorial. The goal is a decision in about five minutes, so you stop burning an evening on a take that was never going to work, or throwing away one that only needed an EQ move.

Check the file format first, because it can reverse the verdict

A waveform with flat tops is the classic sign of a ruined take, and most people stop looking there. That is a mistake, because what the flat top means depends entirely on what format the file was recorded in.

In a 24-bit or 16-bit file, full scale is a hard ceiling. Any part of the signal that went past it was not stored quietly somewhere for later, it was never captured at all. A 32-bit floating-point file does not work that way. It can represent values above 0 dBFS, so if the overload happened in the digital domain after the converter, the peaks may still be sitting in the file above the ceiling, waiting for you to turn the take down.

RØDE's own REAPER setup guide for 32-bit float recording makes the claim plainly: after you pull the level back down, "the previously 'clipped' audio information is still intact." The guide sets REAPER up by going to File, Project Settings, Media, Recording and setting the WAV bit depth to 32 bit FP, and recovers a hot take through Item Properties, Take Properties, Volume/pan, Normalize, with the peak target lowered from the default 0.0 dB to something like -3.0 dB.

What to do right now: right-click the item, open Item Properties, and look at the source file's format. If it says 32-bit float, try lowering the take volume by 6 to 12 dB and listen to the loud syllables again before you conclude anything. If it says 24-bit or 16-bit, the flat tops are real and you are in the unfixable category below.

The honest limit on this

Thirty-two-bit float protects you from overload that happens after the analogue-to-digital converter. It does nothing about overload that happened before it. If your microphone preamp was turned up too far and distorted, or the converter itself ran out of range, the file faithfully stores that distortion and no amount of turning it down recovers anything.

This gets blurred in marketing, so it is worth being precise. Sound on Sound's review of the RØDE NT1 5th Gen explains that the microphone's resistance to clipping comes from its hardware, two A-D converters running in parallel, one aligned so the analogue circuitry cannot reach clipping even at maximum gain. The 32-bit float format is what preserves that protection through the digital chain. The review is equally clear about the failure case: select the mic as a 24-bit source in your DAW with the gain too high and "the recorded signal will be clipped." The format alone is not the protection. The capture chain has to have the headroom first.

The four things that are actually unfixable

1. Real clipping and preamp overload

Once the data is gone, every tool that claims to fix clipping is guessing. The Audacity manual is unusually honest about this in its documentation for Clip Fix, which "attempts to reconstruct clipped regions by interpolating the lost signal" and which it says "is only likely to be effective for repairing lightly clipped audio." That is the whole category in one sentence. Interpolation can plausibly redraw a peak that was flattened for a sample or two. It cannot invent several milliseconds of a vowel.

The test. Zoom in hard on your loudest words. Look at how long the flat sections last and how many of them there are. Then listen on headphones for whether the crunch appears only on the loudest syllable of a line, or on every loud word throughout.

The verdict. A handful of one or two sample flats on a few peaks is usable, and often inaudible once the vocal sits in a mix. Flat tops that last milliseconds, on every loud word, with audible grit on the vowels, is a re-record. If you want to try the salvage anyway before you give up, the method and its limits are in how to fix clipped and distorted audio in REAPER.

2. A noise floor louder than your quiet parts

This is the one people most often misjudge, because a quiet recording sounds fine until you make it loud enough to use. Gain is linear. Turning a vocal up raises everything in the file by the same amount, so the ratio between your voice and the hiss or the room behind it is set at the moment of capture and nothing downstream improves it.

Neumann calculates a microphone's signal-to-noise ratio as 94 dB minus its self-noise in dB-A, and makes the point that matters more for home recording: "even a very quiet recording room will contribute quite a bit more ambient noise than 10 dB-A." In most home setups the limiting noise is the room and the gain staging, not the microphone's electronics.

The test. Solo the take. Find the quietest line you actually intend to keep, not the loudest. Raise the take until that line sits at a normal vocal level, then listen to what came up with it.

The verdict. If the noise is a steady, even bed that sits politely under the voice, you can live with it and treat it later. If it arrives as a wash of room, traffic, or computer fan that competes with the words, the take will not survive being made loud enough to use. If the problem turns out to be level rather than noise, start with how to fix a vocal that was recorded too quietly.

3. Room sound printed onto the take

Reflections off your walls arrived at the capsule at nearly the same time as your voice and were summed into the same signal. There are not two layers in that file to separate. There is one waveform that happens to contain both, and REAPER ships no stock tool that can pull them apart.

The test. Listen to the moment just after you stop speaking or singing. A short, dry click of silence means a controlled capture. An audible tail that rings on, or a boxy honk on sustained notes, means the room is in the recording.

The verdict, and this one has a twist. Mild room sound is normal and you can reduce how distracting it is. Heavy room sound is permanent. But here is the part that matters: re-recording only helps if you change the room. If you sing the same part into the same microphone at the same distance in the same untreated bedroom, you will get the same recording with a different performance on it. Before you re-record for this reason, read how to record vocals in an untreated room and change something physical. If you are keeping the take, reducing room echo on a vocal is about managing the symptom, not removing it.

4. Pitch and timing that are a long way out

Correction tools work, at a cost that rises with how much you ask of them. REAPER's stock ReaTune corrects automatically against a chosen Key and Mode, optionally restricted to a Notes Allowed list, with an Attack Time control that Cockos documents as "the time between the start of a note and when ReaTune starts to correct it if needed." It also offers a Prevent octave shifts option, which exists precisely because harmonics in a voice can make the detector guess the wrong octave. Timing is similar: the play rate control, with Preserve pitch when changing play rate enabled, uses the SoundTouch time-stretching algorithm, and the further you stretch, the more that algorithm shows.

The test. Count the problems. Is it three notes in a chorus, or is it most of the take?

The verdict. Individual notes that are slightly off are a routine fix, and tuning vocals in REAPER with stock ReaTune covers it. A take where you were fighting the key throughout is faster to re-sing than to correct, and heavy correction has its own recognisable sound, which is the subject of why a vocal ends up sounding robotic or over-processed. If the take is good in parts, you may not face an all-or-nothing choice at all. See the next section.

The things that sound unfixable and are not

Almost everything people re-record for belongs here. None of it is a reason to throw a take away.

• It sounds muddy, boxy, or honky. Tone balance, which is what EQ is for. Start with why a home vocal sounds muddy in REAPER.
• It sounds thin, or harsh, or nasal. Same category. Fixable with stock plugins and a careful ear.
• The s sounds are piercing, or the p sounds thump. De-essing and plosive repair are routine and reliable.
• The level jumps around wildly. Ride it with a volume envelope before you reach for heavier compression.
• One word is wrong, flat, or mistimed. You can replace that word rather than the take, with a punch-in that does not announce itself.
• No single take is good all the way through. If you recorded several passes, the answer is usually not to re-record but to build one good take from the best parts of several.

There is also a category that is not a recording problem at all. If the take measures fine and still sounds unflattering to you specifically, the likely explanation is that you are comparing it to the version of your voice you hear through your own skull, which is covered in why your voice sounds worse in the recording than in your head.

A five-minute triage

1. Check the format. Item Properties, source file. 32-bit float means a hot take may be recoverable. 24-bit or 16-bit means the ceiling was real.
2. Zoom in on the loudest word. Flat tops lasting more than a sample or two, on most loud words, is category 1.
3. Raise the quietest line you want to keep. Listen to what comes up behind it. That is category 2.
4. Listen to the second after you stop. A ring or a honk is category 3.
5. Count the pitch and timing problems. A few is a fix, most of the take is category 4.
6. Check whether you have other passes. If you do, comping may beat both fixing and re-recording.

If none of the four categories fired, the take is fine and what you are hearing is an unmixed vocal. That is a different job and a much easier one.

What this will not tell you

It will not tell you whether the performance is any good. A technically clean take of a part you do not believe in is still the wrong take, and no triage catches that.

It will not give you a numeric threshold either. There is no dB figure at which a take becomes unusable, because it depends on how loud and dense the rest of your arrangement is. A noise floor that ruins a sparse acoustic ballad can be inaudible under a full mix. And it will not rescue a take that fails two categories at once: one mild problem is usually survivable, but clipped and roomy is almost always a re-record.

If you do re-record, change something first

The most common mistake after deciding to re-record is to set up exactly as before and get the same file with a different performance. At least one of these should be different.

• The recording level. If you clipped, you need headroom, not luck. Work out your target with how to gain stage vocals in REAPER.
• The distance to the microphone. This changes how much room you capture and how much low end you get, and it is the cheapest variable you have. See how close you should be to the mic.
• The space. Soft furnishings, a different corner, a blanket behind you. Anything that reduces reflections changes the one problem you cannot process away.

And keep the old take. Mute it, do not delete it. A re-record sometimes turns out worse than the thing it replaced, and the point of running this triage rather than guessing is that you now know which parts of the original were actually broken.