mkdiff

Audio similarity checker for two recordings

Compare two recordings with a signal-based similarity score, waveforms and a difference timeline. Check matching exports or edited copies; this is not speaker identification.

Or drop both files here. First: Audio One (original); second: Audio Two (modified). Up to 50 MB each. Check the order below or use Swap Audio before comparing.

Audio One

Supports MP3, WAV, OGG, WebM, AAC, FLAC, M4A

Audio Two

Supports MP3, WAV, OGG, WebM, AAC, FLAC, M4A

How voice, song, and audio similarity is measured

  1. Choose recordings of the same passage and start the comparison. Matching file names and durations do not establish matching sound.
  2. Check timing offset and overall level before interpreting the score. Align a delayed copy; use normalization to investigate a gain change.
  3. Switch between the original and modified recording at the high-difference intervals and review the waveform and spectral information. Describe the observed changes rather than treating a score as proof of identity.

A reproducible similarity check

Compare a file with itself to check the matching baseline, then use Try sample audio to examine a known half-second edit. If two exports sound alike but score differently, check their alignment and whether either file was encoded again.

Can I compare two songs for similarity?

You can compare their decoded recordings, which is useful for alternate mixes or exports of the same passage. This tool does not recognize melodies or judge whether two compositions are similar or plagiarized.

What is a good audio similarity score?

There is no universal passing threshold. A small edit, gain change, timing shift or different encoding can affect the signal match. Decide what changes matter for your recording and confirm them by listening.

Can it tell whether two voices are the same person?

No. Different words, microphones, rooms and speaking timing change the recorded signal. The comparison cannot identify a speaker or authenticate a recording.