Reverse audio search engine

“Reverse audio search engine” sounds like a single box: drop a file, get a match. In practice you are stacking separate systems, and only one of them lives on this site. This page compares the stack so you do not expect a catalogue from a local reverser.

The pieces are:

1. Timeline reverse (this site).
2. Lyric or quote search (any web search box).
3. Acoustic fingerprint / ACR-style apps (Shazam and peers).
4. Humming or melody search (some of those same apps).
5. Human memory and friends.

Each piece fails in a different way. Using the wrong piece first wastes time.

Method A: reverse, then identify

Use this when the recording is already backwards. Reverse restores the order of samples. Identification still happens somewhere else.

On this page: upload MP3, WAV, FLAC, OGG, M4A, or AAC, or record at most 60 seconds. The Web Audio API builds a reversed buffer in the tab. You A/B with Play reversed and Use original. You may preview at 0.5x–2x. You download WAV. No file is copied to our servers. No index is queried.

Then you play the WAV into method B, C, or D. Details for the music case sit on reverse music search. The general pattern is on reverse audio search. Speech is on reverse voice search. Format notes for compressed uploads are on MP3 reverser.

Method B: lyric and quote search

Once you can hear words, type them. Rare lines work better than “I love you.” Add the word “lyrics” if the query is generic. For spoken audio, search names, numbers, and odd phrases.

This method needs intelligible language. It fails on instrumental tracks and on clips that are still reversed. It also fails when you mishear a syllable; reversed audio is a factory for mishearings, which is why you restore direction before you type.

Google’s contribution is text search, not waveform search. See Google reverse audio search for that distinction.

Method C: fingerprint apps (Shazam, SoundHound, ACR-class)

These apps hash short slices of the spectrum in time order and compare the hash to a commercial database. They are good at radio mixes, TV themes, and popular recordings. They are weak on:

• Reversed files (wrong time order).
• Heavy voiceover.
• Extreme covers and karaoke with different arrangement.
• Private voice memos that were never released.
• Sound effects that are not in a music catalogue.

They need a network. This site will not proxy that network call. Hold the phone to your speaker after reverse, or open the WAV in an app that accepts files.

Method D: humming and melody

Humming is for when you have the tune in your head and no decent recording. It is not a substitute for reversing a file. If you only have a backwards MP3, restore it, listen, then hum the forward melody if the fingerprint still misses.

Melody search also collides with songs that share a four-chord pattern. Treat a hummed hit as a candidate list, not a verdict.

Method E: people

Forum posts, group chats, and “name this tune” communities still work when catalogues do not. You still need a forward clip or a written lyric to show them. Posting a still-reversed file usually gets silence.

What a real search engine would require

An actual reverse-audio search engine would store fingerprints or full files, accept uploads, and rank matches. That means servers, licenses, and a privacy policy about retained audio. This project chose not to be that. The browser already knows how to reverse a buffer. Keeping the buffer on the device avoids uploading speech and music you may not want stored.

If you need engine-like behavior, use the dedicated ID apps. They already paid for catalogues. This page stays the reverse step so those apps receive a forward mix.

How to choose in 30 seconds

• Gibberish that might be backwards → reverse here → listen → then C or B.
• Clear song, unknown title → C, then B.
• Tune in your head, no file → D.
• Spoken memo, unknown speaker topic → reverse if needed → transcript → B. Not a speaker-ID product.
• You already know the title and want it backwards → reverse here and export WAV. You do not need an engine.

Technical limits of the reverse step only

One file at a time. Mic capture capped at 60 seconds. Output always WAV. No YouTube URL field. No polarity invert. Preview speed does not change the download. Older phones may struggle with long FLAC or WAV files because decode happens in memory.

If the ID app still fails after a clean reverse, the recording is probably not in the catalogue, or the window you played was too noisy. Trying a different chorus bar helps more than reversing a third time.

Why this page exists

Search consoles show queries for “reverse audio search engine” from people who want one box. Honest documentation beats a fake results list. Use the comparison above, run the local reverse when the timeline is wrong, and send the restored audio to a system that actually searches.