A recording can play successfully and still contain a problem worth fixing before you submit it. One channel might be empty, the recording level might be too high, or a long pause might have slipped into the file. This audio file analyzer gives you a quick view of the decoded signal so you know what to listen for.
Choose a short recording from your device, analyze it and review the channel measurements. Your file is processed locally by your browser. There is no upload, account requirement or microphone request. The results describe the signal; they do not replace listening or the written instructions for your project.
LISTEN / MEASURE / REVIEW
Check a short audio file
Choose a recording to inspect its levels and channels. Your file stays on this device. No account or microphone access is needed.
Your audio measurements
- Duration
- Channels
- Analysis sample rate
- File size
The browser decodes and may resample your recording. The analysis sample rate describes the decoded audio, not necessarily the original file. Use your recorder or audio editor to verify a project's source-format requirements.
| Channel | Sample peak | RMS level | Headroom | Samples ≥ full scale | Low-level windows |
|---|
Listen to your original file
How these measurements are calculated
- Sample peak
- The largest absolute decoded sample, shown as 20 × log10(peak) dBFS. It does not measure inter-sample or true peak.
- RMS level
- The square root of the mean squared samples across the whole channel, converted to dBFS. Pauses affect this average. It is not LUFS or perceived loudness.
- Headroom
- The negative of the sample-peak dBFS value. A negative result means a decoded sample exceeds full scale. An all-zero channel has no finite peak or headroom.
- Samples ≥ full scale
- The count and share of samples with an absolute value of at least 1. These are a reason to listen for distortion, not proof that a recording clipped. Lossy decoding and resampling can change peaks.
- Low-level windows
- The share of channel duration in consecutive 50 ms windows whose RMS is at or below −60 dBFS. A shorter final window is included by its duration. These windows may contain pauses, very quiet speech or background sound; this is not a silence or noise measurement.
How to check an audio file
- Choose a representative section of your recording. Include ordinary speech, a louder phrase and a pause if possible. This tool accepts files up to 20 MiB and 2 minutes.
- Select the file and choose Analyze this file. The browser checks its duration, decodes the audio and measures each channel.
- Look at the channel table. Check whether the channel count makes sense and whether a channel contains only zero samples.
- Play the original file through headphones at a comfortable volume. Listen to the beginning, a louder phrase, a pause and the ending.
- Compare what you hear with the project brief. Keep an untouched original before making any edits.
What the audio measurements tell you
Sample peak is the largest decoded sample in a channel. Its level is shown in dBFS, where zero is full scale. Headroom describes the distance from that sample peak to full scale. A negative headroom value means a decoded sample exceeds full scale. Samples at or above full scale deserve a closer listen, but do not prove that the original recording clipped. Decoding and resampling can affect peaks, and an already distorted recording can have peaks below full scale.
RMS describes the average signal level across a whole channel, including pauses. It helps you compare channels within this recording, but it is not a loudness target. Two files with the same RMS can sound different. This tool does not measure LUFS, true peak, signal-to-noise ratio or speech intelligibility.
Low-level windows show how much of a channel falls into consecutive 50 millisecond sections with RMS at or below −60 dBFS. The final, shorter section is weighted by its duration. These sections can include pauses, very quiet speech or background sound. The percentage is not a count of missing words, a noise score or a reliable measure of silent time.
Why the analysis sample rate can differ from your file
Your browser decodes the file into audio samples and may resample it to the rate used for analysis. The sample rate shown here describes those decoded samples. It does not prove that your original recording was captured or exported at that rate. This tool also does not report the source bit depth or encoded bitrate.
If a project requires a particular source format, check the recorder's export settings or the original file properties in an audio editor. Converting a low-rate recording to a higher rate does not restore detail that was never captured. Follow the brief's instructions for recording, rather than assuming a bigger number means an acceptable file.
Listen for problems the numbers cannot identify
During a pause, listen for fans, traffic, buzzing and room echo. During speech, check for crackling, overloaded syllables, cut-off words and large changes in microphone distance. Listen to both channels if your file is stereo. A channel difference might be intentional, such as separate speakers, or it might point to a setup problem.
The peak overview combines all channels into a visual guide. A flat section tells you where to investigate; it does not explain the cause. A full-scale peak is a reason to listen closely, not an automatic rejection. There is no universal score or pass mark here, because recording tasks can require different devices, environments and formats.
Prepare your next recording
If you hear a problem, make a short test after changing one thing: reduce the recording gain, move away from a noisy appliance or keep a steadier microphone distance. Compare the new test by ear. Do not apply noise removal, compression or normalization unless your project allows it. Processing can change the speech that a collection is meant to capture.
Use the transcription equipment checklist to review your setup, try free transcription practice to build careful listening habits, or use the freelance rate calculator to include recording, checking and correction time in a task estimate. For speech and language project information, read the AI training jobs guide. Joining is free; eligibility and available work depend on each project.
Browser support and file limits
Format support depends on your browser. A valid recording may fail to open if its codec is unsupported or its duration cannot be read. Try exporting a short WAV or MP3 sample in your usual audio editor. The file-size, duration and decoded-audio limits keep this a tool for short checks; very large or unusual multichannel files should be inspected in an audio editor.
The file is not changed or repaired. Use Clear file to remove the current file from this tool. Closing the page also ends the local session.