FLAC vs MP3: lossless compression vs perceptual compression

By Andrew Butson · July 2, 2026 · 7 min read

FLAC and MP3 exist to solve the same practical problem: a raw WAV recording is too big to store or send around comfortably. They take opposite approaches to solving it. FLAC repackages the exact same waveform more efficiently, the way a zip archive repacks a text file: nothing is removed, just written down more compactly. MP3 instead decides which parts of the sound you are unlikely to consciously hear and deletes them outright, which lets it shrink the file far further than any lossless method ever could.

Reversible versus irreversible: that is the entire story here, and it is what should decide which one you reach for.

Two answers to the same problem

Both formats start from the same place: a raw PCM recording that is bigger than it needs to be. FLAC’s answer is to find and remove statistical redundancy without touching the content — it predicts each sample from the ones just before it and stores only the small difference between the prediction and the real value, then packs those differences down further. Because the process is fully reversible, decoding a FLAC file reconstructs the exact original waveform, sample for sample.

MP3’s answer is different in kind, not just degree. Rather than finding redundancy in the numbers, it applies a model of human hearing to the sound itself and removes whatever that model predicts you won’t miss. That is a fundamentally lossy operation: the discarded information is gone, not just compressed. It is exactly what lets MP3 shrink a file far more than any lossless method can.

How FLAC halves the size without losing anything

FLAC roughly halves file size while keeping every sample, which sounds like it should be impossible: lossless compression normally implies "however much redundancy the source happens to contain, and no more." Audio waveforms turn out to contain quite a lot of it: consecutive samples in music are rarely random jumps, they follow the shape of the sound, so a good linear predictor guesses the next sample fairly closely and only the (usually small) error needs to be stored. FLAC further compresses those error values with an efficient entropy coder, and the whole scheme is symmetric — what compression removes, decompression puts back exactly, which is why a FLAC decoder can reconstruct output identical to the original PCM, verified against checksums embedded in the format itself.

How much a given track shrinks still depends on its content: a quiet acoustic recording with a lot of silence and repetition compresses more; a dense, loud, heavily-produced track compresses less, because there is simply less redundancy left to exploit. Either way, nothing about the audio itself is ever at risk — only the file size varies.

How MP3 gets so much smaller

MP3 leans on a psychoacoustic model: quiet sounds masked by louder ones nearby, and frequencies at the edges of human hearing, are reduced or removed because most listeners will not consciously notice their absence. The bitrate setting controls how aggressively this happens — a high bitrate makes conservative, hard-to-notice cuts; a low one cuts much deeper and the shortcuts start to become audible as artefacts on complex material.

This is where MP3 pulls ahead on size. Because it is allowed to remove information rather than merely repackage it, it can go far past what any lossless scheme could ever achieve on the same source, typically a tenth of the original WAV or smaller against FLAC’s roughly half.

Can you actually hear the difference?

Stated plainly and without overclaiming either direction: at high bitrates (256–320 kbps), most people cannot reliably tell a well-encoded MP3 apart from a lossless original in a blind listening test on typical gear. The cuts MP3 makes at that bitrate are specifically chosen to fall below what most ears notice. At lower bitrates (128 kbps and below), the gap opens up, and the difference becomes audible to most listeners on most equipment, especially on complex or percussive passages.

None of that changes what FLAC is for. Even where the audible difference is small or nonexistent, FLAC guarantees it, rather than betting on it, which matters if the file might get re-encoded, remixed, or judged by someone with better ears or gear than the original test used.

The archive-in-FLAC, listen-in-MP3 strategy

The practical workflow that falls out of all this: keep one FLAC per track as the permanent copy (half the size of an uncompressed master with nothing given up) and generate MP3s from it on demand for whatever needs to actually play the file. Phones, cars, older players and streaming all still lean on MP3’s universal support and small size; the FLAC archive means you are never stuck re-ripping or re-recording if you need a fresh, high-bitrate export later, or want to switch to a different lossy format entirely.

This only works in one direction, which is the detail people most often get wrong.

Which direction actually helps

Converting FLAC to MP3 is straightforward and entirely reasonable: it is a normal, one-way lossy export, the same trade-off you’d make going from any lossless source to any compressed one, appropriate whenever the destination is casual or portable listening rather than archiving.

Converting MP3 to FLAC gains nothing. FLAC can only losslessly repackage the audio it is handed — if that audio is already an MP3, the detail MP3’s encoder discarded is already gone, permanently. The result is a larger file holding exactly the same, already-compressed sound, with none of the quality a genuine lossless source would have. If you only have an MP3, an MP3 is what you have; FLAC cannot manufacture back what was never kept.

Frequently asked questions

Does FLAC really keep every sample?

Yes. FLAC’s compression is fully reversible — it predicts each sample and stores only the error, so decoding reconstructs audio that is bit-for-bit identical to the original. That is what "lossless" means in practice, not just in name.

Can I hear the difference between FLAC and MP3?

At 256–320 kbps, most listeners cannot reliably tell them apart in a blind test on typical gear. At lower bitrates the difference becomes audible, particularly on complex or percussive material.

Should I convert my FLAC library to MP3, or the other way round?

FLAC to MP3 is fine whenever you need smaller, more portable files: it’s a standard one-way export. MP3 to FLAC is not worth doing; it gains nothing, since FLAC can only losslessly repackage audio that is already missing the detail MP3 discarded.

Is FLAC as small as MP3?

No. FLAC only halves the size of an uncompressed WAV because it cannot remove any actual information. MP3 gets much smaller (often a tenth of the original or less) precisely because it is allowed to throw detail away.