Audio formats explained: MP3, AAC, WAV, FLAC, and friends

Three families, one decision

Uncompressed audio (WAV, AIFF) stores the raw digital signal exactly as sampled. CD-quality stereo runs about 10 MB per minute — big, but dead simple, which is why recording gear and editing software love it: no decoding, no artifacts, no surprises.

Lossless compression (FLAC, ALAC) shrinks that same signal to roughly half the size the way ZIP shrinks a document: perfectly reversibly. Decode a FLAC and you get back the identical waveform, bit for bit. This is the archival family — all of the fidelity at half the storage, at the cost of spottier device support than the formats below.

Lossy compression (MP3, AAC/M4A, OGG Vorbis, Opus) gets radical: it discards the parts of the sound that human hearing is least able to notice — masked frequencies, inaudibly quiet details — and shrinks files by 80–90%. Done well, at a decent bitrate, the removal is genuinely imperceptible; done aggressively, it produces the swishy, underwater artifacts of a bad stream. Every song you have ever streamed was lossy.

The formats you'll actually meet

Bitrates: where "good enough" actually is

Lossy quality is set by bitrate, in kilobits per second. The scale is worth internalizing because it is where most quality-vs-size decisions happen. For MP3: 128 kbps is serviceable but audibly imperfect on cymbals and complex passages; 192 kbps is the comfortable middle where most listeners stop hearing differences; 256–320 kbps is transparent — indistinguishable from the source in blind tests for nearly everyone, on nearly all material. Newer codecs shift the scale downward: AAC at 192 kbps roughly matches MP3 at 256, and Opus is comfortable lower still.

Spoken word is far more forgiving than music — podcasts and lecture recordings are perfectly pleasant at 64–96 kbps mono, which is why an hour of podcast can weigh less than three pop songs.

The one-way street

The single most important rule in audio conversion: quality only flows downhill. Converting FLAC → MP3 makes a sensible portable copy. Converting MP3 → FLAC makes a file three times larger that sounds exactly like the MP3 — the discarded information does not come back, and the lossless wrapper merely preserves the damage perfectly. The same goes for re-exporting a 128 kbps file at 320 kbps.

The subtler version of the mistake is lossy-to-lossy conversion: M4A → MP3, or re-encoding an MP3 to a different bitrate. Each lossy pass discards a fresh layer of "least noticeable" detail — but what was least noticeable already went in the first pass, so damage compounds (generation loss, the same phenomenon as with repeated video re-encoding). One conversion for compatibility is usually inaudible; make it a habit and the artifacts arrive. Convert from the most original source available, and when you must go lossy-to-lossy, use a generous target bitrate to minimize the new round of loss.

Practical answers to the common cases

• Voice memo / WAV recording to share: WAV → MP3 at 128–192 kbps. Ten times smaller, plays everywhere, nothing anyone will hear was lost on speech.
• FLAC collection onto a picky device or app: FLAC → MP3 at 320 kbps — keep the FLACs as your archive; the MP3s are disposable copies.
• Only need the audio from a video: extract it — a talk, concert, or interview shrinks by 90%+ when the video track goes. Extract to WAV if you plan to edit it; MP3 if you just plan to listen.
• Editing or producing: work in WAV throughout, export to lossy exactly once at the end at 256+ kbps.
• Long-term music archive: FLAC. Half the size of WAV, bit-perfect, open, and you can generate any future format from it without loss.