MP3 320, FLAC, or WAV?

The 20-second answer

Playing out (house party, bar, club rig): MP3 320. In blind listening tests almost nobody can reliably tell a well-encoded 320 kbps MP3 from the lossless original, and on a loud system in a real room the difference is academic. It is also the copy the network almost always has.

Collecting, producing, re-editing: FLAC. Lossless means every re-edit, re-encode and future format change starts from the full original. Once audio has been through a lossy encode, what was discarded never comes back.

WAV: only when a tool demands it. FLAC decodes to the identical audio, at roughly 60% of the size, with proper tags. There is no sound-quality reason to prefer WAV over FLAC.

What each format actually is

WAV is the raw, uncompressed audio: every sample stored as-is. CD quality runs at 1,411 kbps. Big files, and tag support (artist, title, key, BPM) is patchy across players.

FLAC is the same audio, losslessly packed, think of a ZIP of the WAV. Decoded, it is bit-for-bit identical to the original. Around 60% of the size, with first-class tags.

MP3 320 is lossy: a psychoacoustic model discards the parts of the signal your hearing masks anyway, and 320 kbps is MP3 at its ceiling, keeping audio content up to about 20.5 kHz. For scale: children hear up to about 20 kHz, and most adults top out between 15 and 17 kHz. The content a good 320 encode drops sits almost entirely above what an adult ear, let alone a club PA in a crowded room, resolves.

Same track, three sizes

What blind tests actually show

Decades of controlled ABX listening tests, where you do not know which version is playing, point the same way: at 320 kbps, listeners overwhelmingly fail to beat coin-flip odds picking the MP3 from the lossless original on normal music. The exceptions are a handful of known killer samples (harpsichord, applause, castanets) and some trained ears on studio monitors in silent rooms.

A DJ set is the opposite of that lab: a loud PA, a room full of people, crowd noise, and a system tuned for punch rather than analysis. In that setting the difference between a good 320 encode and lossless is not the thing anyone hears. The mix, the levels and the room matter orders of magnitude more.

So why does FLAC exist at all? Because lossless is about what you can do with the file later, not what you hear tonight.

The transcode trap

Converting an MP3 to WAV or FLAC does not improve it. The lossy encode threw data away; converting up just wraps the same audio in a bigger file. A spectrum analyser shows it instantly: a "FLAC" with a hard shelf near 20 kHz was born lossy.

This matters on any peer-to-peer network: some files labelled lossless are secretly transcodes. Honest limit: Cratefox's strict format search matches by file extension; it cannot look inside someone else's FLAC to prove its history. If lossless integrity is critical to you, spot-check downloads with a free spectrum tool like Spek.

The same physics is why re-encoding lossy to lossy compounds the damage: an MP3 of an MP3 loses twice. If a track will ever be edited or re-encoded, that is the real argument for grabbing it lossless today.

One gear note before you go all-FLAC

MP3 and WAV play on effectively every CDJ, XDJ and controller you'll meet in a booth. FLAC does not: Pioneer added FLAC support with the CDJ-2000NXS2 generation, so older club players will not read FLAC from a USB stick (the CDJ-2000nexus still installed in plenty of booths is the classic trap). rekordbox handles FLAC fine at home, but if you play USBs on unknown club gear, MP3 320 is the copy that never surprises you.

Whatever you pick, Cratefox keeps the promise strict: a FLAC run never quietly hands you an MP3. Anything that does not turn up in your format lands in fix & retry, where you can send it back out as another format with one click.