Diagnosis
Grouped by what the rejection actually said. Each one gives the cause, whether software can fix it, and what to change if it cannot.
Match the wording
“RMS is below −23 dB”
The commonest rejection, and usually the result of good instincts: recording conservatively so nothing clips. Adding gain fixes it, but only if the peaks and the noise floor are handled at the same time — raising the level moves both.
“RMS is above −18 dB”
Often the result of running a compressor or a loudness normaliser aimed at podcast or streaming targets, which sit far louder than ACX wants. Turning it down is the whole fix, and it also improves the noise floor.
“Peak values exceed −3 dB”
A plosive on a hard consonant, a hand on the desk, or a mouth click amplified by proximity. A pop filter and four inches of distance prevent most of them at the source; a look-ahead limiter fixes them after the fact without ducking the surrounding words.
“Noise floor exceeds −60 dB”
The usual culprits, in the order they are usually found: a computer or a hard drive in the same room; air conditioning or a fridge; traffic through a single-glazed window; preamp hiss from adding gain at the interface instead of moving closer to the microphone; and a room with hard parallel walls, which raises the floor by reflecting everything, including the noise.
What software can do: pull down the gaps between words so the measured floor drops. That works when your speech is comfortably above the room.
What it cannot do: take the room out from under your voice. If your speech is less than about 18 dB above the noise, nothing produces a clean file — and the settings that make the meter pass make the recording audibly breathe, which reviewers and listeners both notice. At that point the answer is a quieter take: turn off the fridge, record at night, get closer to the microphone, and put something soft on the hardest wall.
“File must be 192 kbps CBR MP3 at 44.1 kHz”
The two common causes are a variable-bitrate export — the default in a lot of software, because it is the better choice everywhere else — and a session recorded at 48 kHz, which is the default for video work. Both are a re-export, and the audio itself is fine.
ACX requires a 192 kbps-or-higher constant-bitrate MP3 at 44.1 kHz. See the ACX audio submission requirements.
“Room tone missing” · “more than 5 seconds of room tone”
Usually caused by editing tightly to the first and last word, or by a noise gate that has removed the tone along with the noise. Add half a second at the head and about a second at the tail, taken from your own recording. Do not paste silence — here is why.
The loop that costs weeks
This is not bad luck. The three level specifications are coupled, so fixing one in isolation moves the other two and the next rejection names a different line.
Raising the level to fix RMS raises the noise floor by the same amount and pushes peaks toward the ceiling. Anything that fixes these properly has to solve them as one problem, in the order the numbers allow: floor first, against the level it has to reach before the gain, then limiting on the signal that actually gets encoded.
And then it has to measure the file it produced, rather than the file it intended to produce.
Stop guessing