Check a Recording Before You Publish It
Drop an audio file and the checker measures the waveform itself: how loud the peaks get, whether anything clipped, the average level, and how much background noise sits under the voice. You get a score out of 100, and under it the actual numbers and what to do about each one.
Run it after you record a podcast, a voiceover or an interview, and before you spend an hour editing something that was broken at the source. Clipping and a loud noise floor are the two problems an edit cannot really fix, so this is the moment to catch them.
Clipping
Counts the samples pinned at the ceiling. Even a small share means distortion that is baked in.
Average level
How loud the recording is over time, with how many dB to add if it is too quiet.
Noise floor and signal to noise
How loud the hiss, hum or room tone is, and how far the voice sits above it.
DC offset
Whether the waveform is shifted off center, a hardware quirk that eats headroom.
What Happens When You Drop a File
There is no AI model guessing here. The checker does arithmetic on the samples in your file, in this order:
1. Decode
Your browser turns the MP3, WAV or other file into raw samples, the numbers that describe the waveform.
2. Mix to mono
The channels are averaged into one, so a problem on only one side of a stereo file still shows up.
3. Scan every sample
It records the loudest sample (the peak), counts the samples at 99.7% of full scale or more (clipping), and averages the rest (RMS level and DC offset).
4. Find the noise floor
It cuts the file into 50 ms slices, measures each one, and takes the quietest 5% of the file as the noise floor.
5. Compare voice and noise
Signal to noise is the average level minus the noise floor, in dB. The bigger the gap, the cleaner it sounds.
6. Grade and score
Each check is marked pass, warning or fail against fixed thresholds, and the weighted results become the score.
Because all of this runs in the browser with the Web Audio API, the recording never leaves your device. Files up to 80MB work, including the audio track of a video file.
What a Report Looks Like
This is the shape of the result. The recording and its numbers are made up for this example.
interview-take-2.wav
Score 73/100
Peak
-2.4 dBFS
Avg level
-26.1 dBFS
Noise floor
-57.6 dBFS
Signal to noise
31.5 dB
Clipping
No clipping. Peak is -2.4 dBFS, with headroom.
Level
A little quiet at -26.1 dBFS. Nudging up about 8 dB gets you to a comfortable spoken level.
Background noise
Signal-to-noise is about 31.5 dB. Usable, but a listener on headphones will hear the room. Noise floor sits at -57.6 dBFS.
DC offset
Waveform is centered, no DC offset worth worrying about.
Format
Mono, 48,000 Hz, 3:12.
The score adds up the five checks. Each one has a weight: clipping 26, level 20, background noise 20, DC offset 8 and format 6. A pass earns the full weight, a warning earns 45% of it, and a fail earns nothing. Read the lines under the score, not just the score. A 73 with clean peaks is an easy fix. A 73 with clipping is a re-record.
The Thresholds Behind Each Check
These are the exact limits the checker uses to mark a check as pass, warning or fail.
| Check | Pass | Warning | Fail |
|---|---|---|---|
| Clipping | Peak below the ceiling, no clipped samples | Peak above -0.3 dBFS | Clipping in more than 0.05% of samples at 99.7% of full scale or more |
| Level (average, RMS) | Average of -23 to -9 dBFS | Average of -30 to -23 dBFS (too quiet) or above -9 dBFS (too hot) | Average below -30 dBFS |
| Background noise (signal to noise) | Gap of 40 dB or more | Gap of 25 to 40 dB | Gap below 25 dB |
| DC offset | Waveform centered | Offset of more than 1% of full scale | Never fails |
Sources, checked 2026-10-09: Audio Quality Checker settings
When the level is low, the checker tells you roughly how many dB to add to reach -18 dBFS, a comfortable average for spoken audio.
How to Fix What It Finds
Clipping
Lower the input gain or move back from the mic, then record again. Clipped peaks cannot be rebuilt.
Too quiet
Raise the gain by the amount the checker suggests, or normalize the file before you publish.
Too hot
Back the gain off a little. A loud average leaves no room for a laugh or a shout.
Noisy
Record in a quieter room first. A noise-reduction pass is the fallback, and the background noise removal tool does it from a file.
DC offset
Apply a DC-removal or high-pass filter in any audio editor. It is a one-click fix.
Wrong format
If a file will not decode, convert it to WAV or MP3 with the MP3 converter and check again.
Use it as a 30-second mic test
Record a short clip
Read a few sentences at your normal distance, with a pause or two.
Drop it in
The pauses give the checker a clean read of the noise floor.
Change one thing
Move the mic, the gain or the room, record again and compare the numbers.
Two numbers do most of the work: the peak tells you about the gain, and the noise floor tells you about the room.
What It Does Not Check
Good at
- Finding clipping and showing how much of the file is affected
- Measuring peak and average level in dBFS
- Estimating the noise floor and signal to noise on speech
- Checking files without uploading them
Not built for
- Taste: it cannot tell if a voice sounds good or the mic suits it
- Loudness in LUFS: use the Audio Loudness Checker for platform targets
- Bitrate and codec: it measures the sound, not how the file is stored
- Fixing the audio: it tells you what to change, it does not edit the file
The signal-to-noise figure is an estimate. It needs quiet gaps to find the noise floor, which speech has and continuous music mostly does not. Trust it most on talk and least on a music bed.
It does not read bitrate or the codec. Bitrate says how much data the file uses, not whether the recording is clean. A 320 kbps MP3 of a clipped recording is still clipped. For codec and per-track bitrate details, a desktop tool such as MediaInfo reads the file’s headers.
Audio Quality Checker vs Other ScreenApp Audio Tools
The names are close, so here is what each one answers.
| Audio Quality Checker | Audio Loudness Checker | Audio Analyzer | |
|---|---|---|---|
| The question it answers | Is this recording clean? | Is it loud enough for Spotify, YouTube or a podcast? | What is in this audio? |
| What it measures | Peak, clipping, average level, noise floor, signal to noise, DC offset | Integrated LUFS and loudness range | Speech, music, instruments, language and sounds, with AI |
| Runs in your browser, no upload | Yes | Yes | No |
| Free account needed | No | No | Yes |
Check quality here first. When it passes, use the Audio Loudness Checker to match a platform’s loudness target, and the Audio Analyzer when you need to know what the audio contains.
After the Check
Trim the dead air
The online audio cutter cuts the start, the end and the long pauses.
Pull audio from a video
The extract audio from videos tool gives you a separate audio file to check and edit.
Get a transcript
The online transcript generator turns the finished recording into text.
Test your mic live
The camera and mic test checks the microphone before you record.
Who Uses an Audio Quality Checker
Podcasters check each raw track before editing, so a clipped guest mic gets re-recorded the same day instead of discovered a week later.
Voiceover artists send a quick test file and the numbers to a client, to show the booth is clean before the session.
Video creators check the audio of a talking-head take before they cut it, because bad sound loses viewers faster than bad picture.
Musicians and producers check a bounced vocal or sample for clipping and DC offset before it goes into a project.
Anyone buying a microphone records the same sentences on the old mic and the new one and compares the noise floor.
