1

00:00:00,000 --> 00:00:02,000

Audio quality checker

Recording quality decides more of a transcript than any software choice made afterwards, and most of it can be measured before you start. This tool decodes your file in the browser and reads the levels off the audio: whether it is clipped, how far the speech sits above the background, and whether the sample rate can carry the detail. It tells you what will cost you words and what will not.

Painted illustration for the audio quality checker: a microphone beside a level meter, clean even waves passing through onto a ruled page

Drop an audio or video file here

Your file is decoded in your browser and never uploaded. The thresholds are rules of thumb, not a grade: a recording flagged here can still transcribe well, and a clean one can still trip on names and accents. How to record so this comes back clean.

How to use it

  1. Drop an audio or video file onto the box below. It is decoded on your machine and never uploaded.
  2. Read the verdict and the findings underneath, worst first.
  3. Fix what is fixable by re-recording, or go ahead knowing where the rough patches will be.

Knowing where the rough patches are is half the job. FastScribe returns a timestamped draft you can check against the audio at exactly those points.

Transcribe it and check the rough spots →

Frequently asked questions

What is clipping and why does it matter?

Clipping is what happens when a recording is too loud for the file to hold: the peaks are cut flat and the shape of the sound is lost. Unlike noise, it cannot be cleaned up afterwards, because the information is not in the file any more. Heavy clipping costs whole words. The fix is to lower the input gain and record again.

What signal-to-noise figure should I aim for?

Roughly speaking, twenty decibels or more between the speech and the background is comfortable, and under twelve is where transcription starts losing phrases rather than the odd word. Moving the microphone closer to the speaker raises that gap more than any filter or noise removal does.

Is a quiet recording a problem?

Quiet by itself is not fatal, since the level can be raised afterwards. What matters is why it is quiet. Usually it means the microphone was far from the speaker, and distance brings room echo and background with it, which is the part that actually costs accuracy.

What does a narrow frequency range mean?

It means the recording carries almost nothing above the telephone band, around 3.4 kHz. Phone calls, voicemail and some video-call recordings are limited that way at the source. It blurs consonants that differ mostly in their high frequencies, like S and F, and nothing can put that detail back afterwards, so it is worth knowing before you judge the transcript.

Why does this not show my file's sample rate?

Because your browser will not tell us. Decoding resamples every file to whatever rate your own audio hardware runs at, so the sample rate available to this page describes your computer rather than your recording. The frequency range is measured from the audio itself instead, which is the better question anyway: a voicemail re-saved at a high sample rate is still a voicemail, and only the content shows it.

Is my audio uploaded to check it?

No. Your browser decodes the file on your own machine and the numbers are worked out there. Nothing is sent anywhere, and closing the tab discards it.