Lossless vs Lossy Audio: The Only Explanation You Need

Every audio format you've ever met (FLAC, MP3, AAC, WAV, Opus, ALAC, the whole alphabet soup) sorts into exactly two piles. Once you understand the two piles, the soup stops being confusing forever. That's the whole promise of this post: one distinction, properly explained, and you'll never squint at a format name again.

The distinction is lossless vs lossy, and it comes down to a single question: when this file was made smaller, was anything thrown away?

The two piles

Lossless formats compress audio without discarding any of it. Decode the file and you get the original recording back exactly, every sample, bit-for-bit. The compression works like a ZIP archive: it finds patterns and redundancy in the signal and describes them more efficiently, which is possible because music is full of predictable structure. Nothing is approximated. Unzip the book, and every word is there.

Lossy formats compress audio by deleting parts of it, permanently. They use a psychoacoustic model, a mathematical description of human hearing, to identify the details you're least likely to notice: quiet sounds sitting right next to loud ones in frequency, soft details in the shadow of a drum hit. Those get discarded, and the file shrinks dramatically. Decode a lossy file and you get an approximation, usually an excellent one, but never the original.

Lossless asks "how can I describe this signal more efficiently?" Lossy asks "how much of this signal can I get away with deleting?"

Both are legitimate engineering. Neither is a scam. But they are doing fundamentally different jobs, and most format confusion comes from not knowing which job a given file was built for.

Sorting the alphabet soup

Here's the soup, sorted:

Lossless:

• FLAC: the open standard, universally supported outside Apple's walled garden, and the sensible default for a music archive. (The full FLAC explainer is here.)
• ALAC: Apple's lossless codec. Same idea as FLAC, same audible result, plays nicely with Apple gear.
• WAV / AIFF, technically lossless, but uncompressed: the raw samples written out at full size. Twice as big as FLAC for the identical sound, with clumsy metadata support.
• WavPack, APE: niche lossless formats. Fine, but with far less support than FLAC.

Lossy:

• MP3: the one that started it all. Old but everywhere; excellent at high bitrates.
• AAC: MP3's successor, a bit more efficient at the same bitrate. What Apple and YouTube use.
• Ogg Vorbis: the open lossy codec of the 2000s; Spotify used it for years.
• Opus, the modern champion: remarkably good even at low bitrates, and what most real-time and modern streaming audio runs on.

A crucial point hiding in that list: the format tells you the floor, not the ceiling. A FLAC file is guaranteed to be a perfect copy of its source, but if the source was bad, the FLAC faithfully preserves bad. And a FLAC transcoded from an MP3 is a lossless copy of a lossy file: MP3 quality wearing a FLAC badge. The pile a file sits in tells you what the compression did, not how good the recording is.

The one-way door

Here's the property that makes this distinction matter more than any other spec on the box: lossy is irreversible.

When an MP3 encoder deletes detail, that detail does not exist anymore. No converter, no AI upscaler, no amount of money spent on cables will get it back. You can convert an MP3 to FLAC and the file will dutifully grow five times larger, and sound exactly, precisely the same, because you can't losslessly preserve information that was already destroyed.

Worse, lossy damage compounds. Convert an MP3 to AAC and the AAC encoder deletes its own set of "inaudible" details, from an already-thinned signal. Each generation stacks artifacts on artifacts. A file that's been through two or three lossy hops can sound noticeably worse than either codec would ever produce alone.

Lossless has none of this. You can decode, re-encode, convert FLAC to ALAC to WAV and back a thousand times, and the thousandth copy is bit-identical to the first. There is nothing to lose.

Lossy is a one-way door. You can walk through it any time you like, so there's never a reason to walk through it early.

This is why the standard advice isn't "lossless good, lossy bad." It's about direction: archive in lossless, derive lossy copies from the archive when you need them, and never let a lossy file be the only copy of something you care about.

When each is the right tool

Because here's the honest part: lossy compression is one of the great engineering achievements in audio, and there are places it's simply the correct choice.

Lossy earns its keep when:

• Space or bandwidth is genuinely constrained. A phone with your whole library on it, at bitrates where you'd need an ABX test to tell the difference. That's lossy doing exactly what it was invented for.
• The link is lossy anyway. Standard Bluetooth re-encodes whatever you send through its own lossy codec. Streaming a pristine source over it still helps (you avoid stacking losses), but the last hop isn't lossless regardless.
• The content doesn't need it. Podcasts and audiobooks are speech; Opus at modest bitrates handles them perfectly.

Lossless earns its keep when:

• It's the master copy. Your archive, your CD rips, anything you intend to keep for decades. Storage is cheap; re-buying or re-ripping your collection is not.
• The file will be edited or converted again. Every future format change should start from a perfect source.
• You want the question closed. With lossless, "did compression change the sound?" has a hard, mathematical answer: no. No listening test required, no doubt to manage.

One trap worth flagging: "lossless" is not the same claim as "hi-res." Lossless promises perfect preservation of the source; it says nothing about whether that source is CD-quality or 24-bit / 96 kHz, or whether the difference between those is even audible. That's a separate (and much more contested) question, which the hi-res audio post takes on directly.

The takeaway

Two piles. Lossless keeps everything and promises you the original back; lossy deletes cleverly and promises you won't notice. Both keep their promises remarkably well. The only mistake you can really make is direction: letting a lossy file become the ancestor of everything else, when it should always be the descendant.

Keep your masters lossless, spend lossy freely at the edges, and the alphabet soup never confuses you again.

And if you're curious what's actually inside your files, open the player and drag some in. It reads the real codec, bitrate and sample rate straight from the file, right in your browser, without a single byte leaving your machine. It's a surprisingly quick way to find out which pile your collection is really sitting in.