FLAC, MP3 or AAC: which format is actually worth the storage
Lossless files are four to six times larger than a good AAC encode. Whether you can hear the difference depends far less on the format than on three things almost nobody checks first.
The format argument has been running since the late nineties and it almost never gets settled, because most of it is conducted without anybody testing anything. So here is the practical version: what the formats actually do, what decides whether you can hear a difference, and how to find out for yourself in about fifteen minutes.
What the three formats are
FLAC is lossless compression. It works like a ZIP file for audio: it shrinks the data, and when you play it back you get every original sample returned exactly. A CD-quality stereo track (16-bit, 44.1 kHz) runs at roughly 1,411 kbps uncompressed, and FLAC typically brings that down to somewhere between 700 and 1,000 kbps depending on the music. Dense, loud material compresses less than sparse, quiet material.
MP3 is lossy. It permanently discards audio information the encoder predicts you won't notice, based on a psychoacoustic model — largely masking, the effect where a loud sound hides a quieter one nearby in frequency or time. It's old, it's universally supported, and modern encoders are far better than the ones that gave MP3 its reputation. LAME at V0 (variable bitrate, averaging around 245 kbps) or 320 kbps constant is about as good as the format gets.
AAC is also lossy, and is broadly the better-engineered successor. At the same bitrate it generally outperforms MP3, with the gap widest at lower bitrates. At 256 kbps, AAC is transparent for most listeners on most material — meaning they cannot reliably distinguish it from the source under controlled conditions.
The storage maths
| Format | Typical bitrate | Approx. per 4-min track | 1,000 albums |
|---|---|---|---|
| FLAC (16/44.1) | ~850 kbps | ~25 MB | ~250 GB |
| AAC 256 | 256 kbps | ~7.7 MB | ~77 GB |
| MP3 320 | 320 kbps | ~9.6 MB | ~96 GB |
| AAC 128 | 128 kbps | ~3.8 MB | ~38 GB |
Assuming roughly ten tracks per album. On a 256 GB phone, a lossless library of any size becomes a real constraint; a 256 kbps AAC library mostly doesn't.
The three things that matter more than the format
1. Your playback chain
The difference between throwaway earbuds and a competent mid-range pair is not subtle, and it is larger than the difference between AAC 256 and FLAC on the same equipment. If you are listening on something that can't resolve detail, feeding it more detail achieves nothing. Spend on the transducer before you spend on storage.
2. The master, not the format
This is the one that catches people out. A lot of perceived quality differences between services or releases are differences in mastering, not encoding. The same album can exist in a heavily compressed, loudness-maximised version and a more dynamic one, and the gap between those two masters dwarfs any codec difference. If a FLAC rip sounds better than a stream, it is often because it came from a different, better master — not because of the file format.
3. Where you listen
On a train, in a car, in an open-plan office, the ambient noise floor sits well above the level at which lossy encoding artifacts live. Transparency is much easier to achieve in noisy conditions. A quiet room with good speakers is the only setting where the format question becomes interesting for most people.
The Bluetooth catch. Standard Bluetooth cannot carry a lossless stream. Whatever file you feed it, your phone re-encodes to a Bluetooth codec — SBC, AAC, aptX, LDAC — before transmission. If you listen on wireless headphones, playing FLAC means decoding lossless audio and then immediately lossily re-encoding it. You keep the storage cost and lose the benefit.
Test it on your own ears
Opinions about audio are cheap because almost nobody tests blind, and sighted listening is unreliable — knowing which file is playing changes what you hear. A proper test is an ABX test: you're given A, you're given B, then you're given X at random and asked which one it is. Do it fifteen or twenty times and the result is statistically meaningful.
- Pick three tracks you know intimately, ideally with different textures — something sparse and acoustic, something dense, something with a lot of cymbals or applause (transient and noise-like content is where lossy encoders struggle most).
- Encode each from your lossless source to AAC 256 and to AAC 128. Keep the lossless original.
- Use an ABX tool. On desktop, foobar2000's ABX comparator plugin is the long-standing option; there are browser-based ABX tools too.
- Run at least sixteen trials per comparison. Getting twelve or more right suggests you genuinely hear a difference. Getting eight or nine out of sixteen is chance.
Most people who do this honestly are surprised. A common outcome is reliably distinguishing 128 from lossless, and failing to distinguish 256 from lossless on the same equipment.
So what should you actually do
Keep lossless as your archive. If you rip CDs or buy downloads, store FLAC on a hard drive at home. It is your master copy. You can always produce a lossy version from a lossless original; you cannot go the other way and recover what was thrown out.
Use lossy on portable devices. AAC at 256 kbps for almost everyone. It's transparent in the conditions you'll actually be in, and it triples what fits on the device.
Ignore bitrate above 320 kbps for lossy. The returns flatten hard. If you want better than transparent AAC, the answer is lossless, not a bigger MP3.
Don't re-encode between lossy formats. Converting an MP3 to AAC applies a second round of loss on top of the first. If you only have lossy files, leave them as they are.
A note on "hi-res"
Files above CD quality — 24-bit, 96 kHz and up — are a separate question from lossless versus lossy. The extra bit depth extends dynamic range well beyond what any domestic listening environment can use, and the extra sample rate extends frequency response above human hearing. Controlled listening tests have generally struggled to show a reliable audible benefit over CD-quality playback. Where hi-res releases do sound better, it's frequently because they were mastered separately and more carefully — which brings us back to point two. That's worth paying for. The sample rate on the label mostly isn't.