Measure your audio's noise floor, room tone and signal-to-noise ratio in dBFS — with an instant clean/noisy verdict.
AI-powered audio review & delivery for voice production teams
Drop your audio file here or click to browse
MP3, WAV, FLAC, AAC, M4A, OGG
This is a meter, not a fixer.
Too much hiss? Reduce it with the free Background Voice Remover. Want the full 6-point check? Run the Voice Quality Analyzer. Checking levels? Use the LUFS Meter.
Upload audio to measure its noise floor, room tone and signal-to-noise ratio
Room tone is the sound of your recording space when nobody is speaking — the ambient bed of air, equipment hum and street noise that is always present. The noise floor is the level of that room tone, measured in dBFS (decibels relative to full scale). The lower (more negative) the number, the quieter and cleaner your background. A recording at −62 dBFS has a far quieter floor than one at −38 dBFS.
Capturing a few seconds of clean room tone at every session is standard practice: editors paste it under cuts and breaths so the silence between words matches the rest of the take instead of dropping to a jarring digital nothing.
This tool reads the quietest sustained passages of your file and reports their RMS level in dBFS. It also reports your program level (the overall RMS of the whole file) and your signal-to-noise ratio (SNR) — the gap, in decibels, between your dialogue and the noise floor beneath it. A −18 dBFS program over a −62 dBFS floor gives an SNR of about 44 dB.
There is no single legal limit, but these are the practical targets professionals deliver to:
| Noise floor | Verdict | What it means |
|---|---|---|
| −60 dBFS or lower | Excellent room tone | Studio-quiet; deliver with confidence |
| −60 to −50 dBFS | Good | Clean enough for spoken-word delivery |
| −50 to −40 dBFS | Audible hiss | Hiss is noticeable; reduce it before delivery |
| Above −40 dBFS | Noisy | Background noise is intrusive; clean it up |
For spoken voice, aim for a signal-to-noise ratio of around 60 dB or more. If your SNR is low even though your floor looks acceptable, your dialogue itself is recorded too quietly — record hotter or move closer to the mic rather than amplifying noise in post.
For spoken voice, aim for a noise floor of −60 dBFS or lower — that is studio-quiet and safe to deliver. Between −60 and −50 dBFS is still good for podcasts and voice-over. Once the floor climbs above −50 dBFS the hiss becomes audible, and above −40 dBFS it is intrusive enough that most clients will reject the file. Upload your recording above to see exactly where it sits.
Room tone is the ambient sound of your recording space with no one speaking — the quiet bed of air, equipment and environment that is always present. Its level is your noise floor. Recording 10–30 seconds of clean room tone at every session gives editors material to fill gaps and smooth cuts so the silences match the rest of the take.
Aim for a signal-to-noise ratio of about 60 dB or more for clean spoken voice. SNR is the gap between your program level and your noise floor, so you can improve it either by lowering the noise floor or by recording your dialogue at a healthier level. A low SNR with an otherwise quiet floor usually means the voice was recorded too softly.
Capture a quieter floor at the source first: treat the room, turn off fans and air conditioning, move closer to the mic, and use a mic with low self-noise. What is already recorded can be reduced afterwards with a dedicated noise-reduction pass — the free Background Voice Remover in the Related Tools is built for exactly that.
dBFS means decibels relative to full scale: 0 dBFS is the loudest a digital file can go, so every level you record is a negative number. A plain dB figure describes a difference between two levels — your SNR, for example, is the dB gap between your program and your noise floor. This tool reports the noise floor, program level and peak in dBFS, and the SNR as a dB difference.
VoiceDeck adds AI-powered audio & video review and delivery for your whole team — so every file ships in spec, automatically.
