Is FLAC Really Lossless?
The definition
"Lossless" has a precise meaning: the decoded audio is bit-for-bit identical to what was encoded. FLAC — the Free Lossless Audio Codec, maintained by the nonprofit Xiph.Org Foundation — meets that definition. Compress a WAV file to FLAC, decompress it, and you get the original samples back exactly. Nothing is approximated, rounded, or discarded. The FLAC documentation is the primary reference for the format and its tools.
This is the same sense in which a ZIP archive is lossless for documents. The question is not whether quality survives — it does, by construction — but how much smaller the file gets.
How FLAC shrinks audio without losing it
FLAC works by prediction. Audio samples are highly correlated — each one is usually close to a value predictable from its neighbors — so the encoder stores a simple mathematical predictor plus the small differences (residuals) between prediction and reality, encoded efficiently. Because music varies in how predictable it is, so does the savings: sparse acoustic material compresses further than dense, noisy material. As a rough characterization, FLAC files commonly come out around half to two-thirds the size of the uncompressed original — the format's documentation covers the encoding levels that trade encoding time against compression.
FLAC also stores a checksum (an MD5 signature) of the original audio inside the file, so decoders can verify that decoded output matches what was encoded — a built-in integrity check lossy formats have no equivalent of.
How lossy formats differ
MP3, AAC, Ogg Vorbis, and Opus take the opposite bargain: they achieve much smaller files by permanently discarding detail chosen to be hard to hear. The selection leans on psychoacoustics — notably that loud sounds mask quieter ones at nearby frequencies, and that human hearing spans a limited frequency range at all. The physics of frequency and amplitude behind those limits is laid out well in the National Park Service's acoustics explainer.
Two consequences follow:
- Lossy is one-way. Converting MP3 back to WAV or FLAC does not restore what was discarded; it only wraps the already-reduced audio in a bigger container.
- Lossy generations stack. Re-encoding lossy audio to another lossy format discards detail again. Archives and masters are kept lossless precisely to avoid this.
When each makes sense
Lossless suits archiving, editing sources, and any file you may re-encode later. Lossy suits delivery where space or bandwidth is constrained and no further processing is planned. The size difference is a bitrate question — a lossless CD-quality stream runs around 1,411 kbps before compression, while common lossy encodes run far lower; the bitrate converter on the home page translates any rate into megabytes per hour. For what the underlying sample numbers mean, see sample rate and bit depth.