AAC vs MP3: which is better, and when it actually matters

Both codecs work the same way in outline. They split the signal into frequency bands, model which parts your ear will mask anyway, and spend bits only on what is left. AAC came four years later and improved nearly every stage: a pure MDCT filter bank instead of MP3's hybrid one, finer control over block sizes during transients, and smarter joint stereo coding.

In practice that shows up at the bottom of the bitrate range. Public listening tests have put AAC comfortably ahead of MP3 at 64 and 96 kbps, which is why podcasts, audiobooks and mobile streaming standardised on it. Climb to 192 kbps and the difference narrows to something most people cannot pick out on ordinary headphones. By 256 kbps both codecs are transparent for practically every listener and every piece of music — a well-encoded 256 kbps MP3 is not a compromise.

Two caveats worth knowing. The encoder matters as much as the codec: a modern LAME MP3 beats a bad AAC encoder, and AAC encoders vary far more between implementations than MP3 encoders do. And none of this applies to files that already exist — efficiency is a property of the encode, not of the file sitting on your disk.

A constant-bitrate file's size is bitrate times duration, so a 192 kbps AAC and a 192 kbps MP3 of the same track are the same size — around 5.8 MB for four minutes. Anyone claiming AAC files are inherently smaller is comparing different bitrates.

The useful version of the claim is that AAC lets you choose a lower bitrate for the same perceived quality. Rough equivalences for music:

• 128 kbps AAC (≈3.8 MB per 4-minute track) sits somewhere around a 160–192 kbps MP3 (4.8–5.8 MB).
• 256 kbps AAC (≈7.7 MB) matches or beats a 320 kbps MP3 (9.6 MB) — the pairing Apple settled on for iTunes Plus.
• For speech, 64 kbps AAC holds up where a 64 kbps MP3 does not, which is worth roughly half the storage on a long audiobook.

That saving is only available when you encode from a lossless master — a CD rip, a WAV, a FLAC. Re-encoding a finished MP3 to a lower-bitrate AAC shrinks the file and damages the audio; it does not buy you AAC's efficiency retroactively.

MP3's last patents expired in 2017, so any manufacturer can ship a decoder for nothing. AAC decoding still sits inside a licensing programme. Combine that with twenty-five years of head start and you get the situation most people actually run into: modern phones, computers, browsers and smart speakers handle both formats, and a long tail of hardware handles only MP3.

The usual offenders, in rough order of how often they come up:

• Car head units, especially anything reading from a USB stick or SD card rather than streaming over Bluetooth.
• DJ and karaoke software, plus older CDJ hardware, where MP3 and WAV are frequently the only supported imports.
• Standalone MP3 players, running-watch storage, and the audio slots on gym equipment.
• Embedded and industrial gear — PA systems, in-store players, telephony prompts, museum guides.

There is a subtler compatibility point too. When you play an MP3 over Bluetooth, the phone decodes it and re-encodes it into whatever A2DP codec the headphones negotiate — usually SBC or AAC. MP3 is technically an optional A2DP codec but almost nothing implements it, so a “pure MP3” wireless path does not really exist.

Three situations cover almost everything:

• Ripping or encoding from a lossless source. Choose AAC at 256 kbps. You get transparent quality in a smaller file, correct gapless playback, and metadata that lives in the container rather than in tags stapled to the front.
• You already have a library. Leave it alone. Converting in either direction costs quality, and the format you have is the format your player already copes with.
• Something refuses to play a file. Convert that file — and it is almost always AAC that needs converting to MP3, not the reverse.

If you are archiving something you cannot re-acquire, neither codec is the answer. Keep a FLAC or WAV master and generate lossy copies from it whenever a device demands one; that way every conversion is a first-generation encode.

Going from AAC to MP3 is the common direction, because it is the one that fixes a device that will not play your file. Drop .aac, .m4a, .m4b or .m4r files on the AAC to MP3 converter and they are decoded and re-encoded at 128 to 320 kbps on your own device — no upload, no signup, 20 files per batch. If your files came out of iTunes, Apple Music or iPhone Voice Memos, the M4A to MP3 page covers that flavour specifically. Pick 256 or 320 kbps for music you care about and 128 kbps for speech.

Going the other way needs a desktop tool: browsers can decode AAC but cannot encode it, so ffmpeg, the Music app or afconvert do that job. The guide on converting MP3 to AAC walks through each of them and explains why the transcode is usually not worth doing.

Whichever direction you go, work on copies. Both formats are lossy, so every conversion is a one-way door — the detail an encoder discards is not recoverable by converting back. The about page describes the exact pipeline this site uses.