FLAC vs AAC: which should you use?

FLAC keeps every bit of the original. AAC throws part of it away to save space. The choice is not really about sound quality — it is about what you plan to do with the file next.

In 30 seconds

Keep FLAC whenever the file is your master, your archive, or something you might convert again later — it is the only version you can re-encode from without losing anything. Use AAC when you need small files that play everywhere Apple devices and mobile already expect. At 256 kbps and above, AAC is transparent to most listeners on everyday gear. What you should avoid is low-bitrate AAC: in a 2001 community blind test, AAC finished last of five encoders at 64 kbps, with a rank sum of 23.0 against 49.0 for the winner. (That was 2001-era AAC — see the caveat below before you read it as a verdict on modern encoders.)

What each format actually is

FLAC is lossless. It compresses the audio the way a ZIP file compresses a document: nothing is discarded, and decoding gives you back a bit-for-bit identical copy of the source. Typical result is roughly half the size of the original CD audio.

AAC is lossy. It uses a psychoacoustic model to decide which parts of the signal you are least likely to notice, then permanently discards them. Typical sizes are a small fraction of the FLAC — which is the entire point.

The asymmetry matters more than the file sizes: FLAC to AAC is a one-way door. You can go from FLAC to AAC as many times as you like, but you cannot go back. What was thrown away is gone.

Where AAC wins

Where FLAC wins

What the 2001 blind tests measured at 64 kbps

Five encoders were rated across twelve samples using the ITU-R five-point scale (5.0 imperceptible, 4.0 perceptible but not annoying, 3.0 slightly annoying, 2.0 annoying, 1.0 very annoying). Each sample was ranked best-to-worst and the ranks summed:

Encoder Rank sum (12 samples)
mp3pro49.0
Ogg Vorbis (quality-based)44.0
Ogg Vorbis (64 kbps target)40.0
WMA824.0
AAC23.0

A Friedman non-parametric analysis was then run on the same data. The finding was more nuanced than the raw ranking suggests: mp3pro took first place overall, but it was not significantly better than either Ogg Vorbis entry. WMA8 and AAC, on the other hand, were clearly at the bottom.

Read this before you judge modern AAC by those numbers

The AAC encoder in that test was a 2001 build of a Fraunhofer-designed implementation running at 64 kbps — a bitrate that no one recommends for AAC today. Modern AAC encoders (Apple's, and the open FDK-AAC) are substantially better at the same bitrate. So the honest reading is:

The 128 kbps test in the same series made a related point: AAC was run in constant-bitrate mode because its VBR mode encoded too high unless the bandwidth was pushed below 16 kHz. On one album used for calibration (Radiohead, The Bends) AAC measured 131 kbps, against 128 for Lame MP3 and 135 for MPC.

Where this data stops being valid — quoted as written (ff123, 2001)

"At the higher bitrates, my listening tests will mean less and less. At 192kbs and above, for example, I won't be able to properly tell you what codec sounds best, since most codecs are already removing the vast majority of midrange artifacts, making it mostly a contest of which codec has the best high-frequency response. For these listening tests, you'll have to rely upon somebody else."

He tightened it further two weeks into February 2001: "even at 128 kbit/s, I may lack the high-frequency ability necessary to properly evaluate the quality of an mp3 codec."

The practical rule

Try it yourself instead of trusting a number

Every figure on this page comes from someone else's ears, on someone else's equipment, in 2001. Yours are the only ones that decide anything. Because conversion here runs in your browser, you can make an AAC or MP3 copy at two different bitrates, play them back-to-back, and check whether you can actually tell — no upload, no cost, no account. If you want to do it properly, the same test organisers used an ABX rule of thumb: 12 correct identifications out of 16 trials counts as really hearing a difference at 95% confidence. Below that, treat the two files as identical for your purposes.

Open the FLAC to MP3 converter →