Audio Comparison — A/B & ABX Blind Test

A free in-browser tool that compares two audio files on objective acoustic metrics. Load File A and File B for a side-by-side breakdown: codec, duration, peak and RMS, integrated LUFS and loudness range, true peak and clipping, spectral cutoff, stereo width, phase correlation and polarity. It aligns the two recordings, reports whether they are bit-identical (with a per-second map of where they differ), and offers a blind ABX test when you want to know if the difference is audible at all.

How to compare two files

Drop your first file into the File A slot (left).

Drop your second file into the File B slot (right).

Press Analyze. Both files decode in parallel; results appear within seconds for typical track lengths.

Read the side-by-side table; differences worth attention are highlighted under "Notable differences".

See exactly which file sounds better — and why

A File A

Drop file A here or click to browse

MP4, WEBM, MOV, MP3, WAV, M4A, OGG, FLAC, AAC, M4B, WMA, AIFF, OPUS, CAF, MKV, AVI, WMV, FLV, M4V, 3GP, TS, MTS, M2TS, VOB, MPG, MPEG, OGV ·

—

B File B

Drop file B here or click to browse

MP4, WEBM, MOV, MP3, WAV, M4A, OGG, FLAC, AAC, M4B, WMA, AIFF, OPUS, CAF, MKV, AVI, WMV, FLV, M4V, 3GP, TS, MTS, M2TS, VOB, MPG, MPEG, OGV ·

—

Decoding both files…

—

Upload two files and press Analyze to see the side-by-side comparison.

Metric	A	B

How is the verdict computed? (formula + references)

Each file is analyzed across 8 acoustic axes and assigned per-axis scores [0..1]. The verdict is the weighted sum. Edge cases (bit-identical PCM, low-correlation different content) override the score.

Formula

score_X = Σᵢ wᵢ · axisᵢ(X), Σᵢ wᵢ = 1.0, winner = argmax(score)

Per-axis weights

Resolution (0.18) — lossless > lossy; honest format outscores upsampled fake-Hi-Res
Spectral integrity (0.22) — higher cutoff Hz indicates a more complete spectrum (less lossy ceiling)
Dynamic range LRA (0.18) — higher LRA in the -23..-9 LUFS band — NOT "louder = better"
True-peak safety (0.08) — closer to -1 dBTP wins; > 0 dBTP penalised (intersample clipping risk)
Loudness target (0.10) — LUFS-I closer to the streaming target (-14 default) scores higher
Codec quality (0.10) — lossless > high-bitrate VBR > CBR 320 > CBR 192 > CBR 128
Stereo integrity (0.06) — width > 0.05 with positive phase correlation; mono-summed penalised
Bit-perfect diff (0.08) — bit-identical PCM forces "indistinguishable" verdict regardless of other axes

References

ITU-R BS.1770-4 — algorithms to measure audio programme loudness and true-peak level
EBU Tech 3341 — "EBU Mode" metering for BS.1770
EBU Tech 3342 — Loudness Range (LRA) descriptor
AES/AC-3 § 5.3 — ABX double-blind comparison protocol

This is a heuristic score, NOT a statistical confidence interval. Treat as a triage tool — for legally-binding forensics, consult a professional.

Process

Features

Side-by-side, not single-file

Most audio-analysis tools inspect one file. This one is built around the comparison itself — every metric you read is paired across A and B, with the difference surfaced as the headline.

Magic-byte container sniff

Identifies the real format from the first 64 bytes regardless of extension. Catches files that say .mp3 but are M4A, OGG-as-WAV, fake-extension renames.

Instant, no render queue

Decoding and analysis happen on the page itself. Drop in two 5-minute tracks and the full LUFS / spectrum / alignment report typically renders in 5–15 seconds on a 2023 laptop — no queue, no "your file is #87 in line", no recurring subscription.

Verdict, not just a table

The tool does more than put metrics side by side. It aligns the recordings, compares the PCM and says it plainly: either "bit-identical", or what share of samples differs ABOVE the quantisation noise and by how many dB at most. When two files really are the same, it says so; when the difference is real but tiny, you see it as a number instead of guessing. For the arguable cases there is a blind ABX test right next to it.

FAQ

Why two slots instead of one?

Audio comparison is a different mental model than single-file analysis. You are not asking "what is this file" — you are asking "which of these two is better, and how?". The whole UI is built around that question.

Does louder mean better?

No. A file mastered to -7 LUFS will be louder than one at -14 LUFS, but the louder one usually trades dynamic range for volume (the so-called "loudness war"). When the full verdict engine ships, dynamic range carries more weight than absolute loudness.

What if my two files are identical?

You will see no differences in the basic table and an "indistinguishable" note. When the sample-level diff lands you will see "bit-identical" if the underlying PCM data is the same, even if the containers differ.

Will you support more than two files?

Maybe. Three-way and N-way comparison is on the roadmap once the two-file flow is validated in production.

💡 Want us to improve this tool just for you?

We can! Just send us a quick message with your idea. If you'd like to discuss it in detail, leave your email and we'll get back to you. You can stay anonymous.

How do you rate this tool?

5 (2 ratings)

Thank you for your rating!

Change my rating

Want to leave a comment?

Want to share more? Leave a comment!

Thank you! Your comment will appear after moderation.

Who is this tool for?

Published 22 May 2026 Updated 03 Jun 2026 Author: The Supreme