Four formats cover almost all everyday audio. They differ on one axis above all: whether they throw information away, and how cleverly.
Audio formats generate more confident misinformation than almost any other technical topic. The underlying distinctions are actually straightforward: some formats store sound exactly as captured, some store it exactly but more compactly, and some discard parts they judge inaudible. Everything else follows from that.
The starting point: what digital audio is
Sound is a continuous pressure wave. Digitising it means measuring amplitude at regular intervals — the sample rate — and storing each measurement with a certain precision — the bit depth. CD audio uses 44,100 samples per second at 16 bits per sample, in stereo. That produces roughly 1.4 megabits per second, or about 10 MB per minute.
The sample rate determines the highest frequency that can be represented: half the sample rate. At 44.1 kHz that ceiling is around 22 kHz, comfortably above the roughly 20 kHz limit of excellent young human hearing. Bit depth determines dynamic range — 16 bits gives about 96 dB, which exceeds the noise floor of most listening environments.
WAV: uncompressed
WAV stores raw samples with essentially no compression. It is the digital equivalent of a master tape: exact, universally readable by editing software, and large.
- Perfect fidelity — nothing is removed or approximated.
- No decoding overhead, so it is ideal for editing and processing.
- Roughly 10 MB per minute of stereo CD-quality audio.
- No standardised metadata support, so tagging is inconsistent.
Use WAV for recording, editing, and mastering. Do not use it for distribution or for filling a phone, where the size is pure waste.
FLAC: lossless compression
FLAC compresses audio losslessly, typically to 50 to 60% of WAV size. Decoding reproduces the original samples bit for bit — it is mathematically identical, not merely similar.
- Bit-perfect reconstruction, verifiable by checksum.
- Around half the size of WAV.
- Excellent metadata support for tags and cover art.
- Open and royalty-free.
- Larger than lossy formats and less universally supported on consumer hardware.
FLAC is the right choice for archiving a music collection, distributing high-resolution releases, and storing masters where space matters. It is the format for keeping, not necessarily for carrying.
MP3: the lossy standard
MP3 discards audio information using psychoacoustic modelling — exploiting the fact that a loud sound masks quieter sounds at nearby frequencies, and that human hearing is far less sensitive at frequency extremes. What is discarded is chosen to be least perceptible, not least important mathematically.
- Universal compatibility. Essentially every device made in the last twenty-five years plays MP3.
- Roughly one tenth the size of WAV at 128 kbps, one seventh at 192 kbps.
- Lossy and generational — re-encoding degrades quality each time.
- Less efficient than newer codecs at the same bitrate.
Practical bitrate guidance: 320 kbps is transparent for nearly all listeners; 192 kbps is very good for casual listening; 128 kbps is acceptable for speech and podcasts; below 96 kbps, artefacts become obvious on music.
AAC: the better lossy option
AAC is the successor to MP3 and is the default in Apple ecosystems, most streaming services, and video containers. At any given bitrate it generally sounds better than MP3, with the advantage most pronounced at lower bitrates.
- Better quality than MP3 at the same bitrate, particularly below 192 kbps.
- Widely supported on modern devices, though not quite as universally as MP3.
- Handles high frequencies and transients more cleanly.
- Still lossy, with the same generational degradation concern.
AAC at 256 kbps is transparent for practically everyone. AAC at 128 kbps is roughly comparable to MP3 at 192 kbps.
Choosing by purpose
- Recording and editing: WAV. Work from uncompressed material and export lossy copies only at the end.
- Archiving a collection: FLAC. Bit-perfect, half the size, well tagged.
- Music on a phone: AAC at 256 kbps, or MP3 at 320 kbps if compatibility is a concern.
- Podcast distribution: MP3 at 96 to 128 kbps mono. Speech does not need stereo or high bitrates, and universal playback matters more than fidelity.
- Audio in a video file: AAC, which is the standard companion to modern video codecs.
- Sending a voice note or clip by email: AAC or MP3 at modest bitrate — keep the attachment small.
Myths worth discarding
"Converting MP3 to FLAC improves quality"
It does not. Information discarded during MP3 encoding is gone permanently. Converting to a lossless format merely stops further loss and produces a much larger file containing exactly the same audio.
"Higher sample rates always sound better"
Above 48 kHz the additional frequency range is beyond human hearing. Higher rates have legitimate uses in production, where processing headroom matters, but for playback the benefit is not audible.
"Lossy audio is always noticeably worse"
In controlled listening tests, high-bitrate AAC and MP3 are indistinguishable from lossless for the large majority of listeners on the large majority of equipment. The audible difference lives at low bitrates, not at the top of the range.
The general rule
Keep masters lossless, distribute lossy, and encode only once from the highest-quality source you have. That single discipline prevents nearly every audio quality problem people encounter. If you also work with documents and images, the same logic applies there — see our guide to file compression for the broader principle, and the Image Compressor for its visual equivalent.