How Compression Affects Audio Quality and Size

Why Audio Files Get So Big in the First Place

Before we talk about shrinking audio, it helps to see the scale of the problem. A CD-quality recording captures 44,100 samples every second, each sample described with 16 bits of precision, across two channels. Multiply that out and you get roughly 1,411 kilobits per second — about 10 MB for every minute of music. A three-minute track lands at around 30 MB, and an hour-long concert recording can comfortably pass 600 MB.

That raw format, usually stored as WAV or AIFF, is wonderfully faithful and wonderfully wasteful. It describes the waveform completely, including enormous amounts of information your ears simply cannot detect. Compression exists to find that surplus and remove it, and the way a codec does so determines whether the result is indistinguishable from the original or a thin, swishing shadow of it.

Lossless Compression: Smaller, Identical

Lossless codecs such as FLAC and ALAC work like a zip file for audio. They look for repetition and predictability in the waveform and store it more efficiently, then rebuild the original signal perfectly on playback. Nothing is discarded.

• Typical savings: 40–60% of the original file size.
• Best case: sparse material like solo piano or spoken word can shrink dramatically.
• Worst case: dense, loud, heavily limited mixes may only save 25–30%.

If you are archiving, editing or mastering, lossless is the sensible default. Storage is cheap, and a lossless master lets you transcode to any format later without stacking damage.

Lossy Compression: Removing What You Won't Miss

Lossy codecs — AAC, MP3, Vorbis, Opus — take a bolder approach. They use a psychoacoustic model of human hearing to decide which parts of the signal can be thrown away without an obvious penalty. Two principles do most of the work:

• Masking: a loud sound temporarily hides quieter sounds close to it in frequency or time, so those quiet details can be dropped.
• Threshold of hearing: content below a certain level at a given frequency is inaudible anyway, so it is discarded outright.

What remains is encoded with a limited budget of bits, allocated where the ear is most sensitive — typically the low midrange. At 256 kbps or above, most listeners on most systems cannot reliably tell the difference from the original in a blind test. Drop to 96 kbps and almost everyone can.

What Too Much Compression Actually Sounds Like

Artefacts are rarely obvious at first. They creep in as small tell-tales that become grating once you notice them:

• Smearing on cymbals and hi-hats, which lose their metallic shimmer and turn into a watery swish.
• Pre-echo: a faint ghost of a drum hit or consonant arriving just before the real one, caused by the encoder's block processing.
• Warbling or bubbling on sustained notes, especially piano and strings.
• Collapsed stereo image, where reverb tails and wide instruments pull towards the centre.
• Pumping on complex, loud passages, as the bit budget runs out and the encoder starts cutting corners.

These are worst on exactly the material you care about most: quiet, dynamic, detailed recordings. Loud, dense pop masters hide artefacts well because the masking is already generous.

Choosing Bitrates Without Guesswork

A few practical starting points, assuming a modern codec such as AAC or Opus:

• Spoken word and podcasts: 64–96 kbps mono is perfectly clear.
• Casual listening, phone speakers, commuting: 128–160 kbps stereo is acceptable, though artefacts appear on good headphones.
• Serious listening: 192–256 kbps, or variable bitrate around quality level 4. This is transparent for the vast majority of people.
• Classical, acoustic and dynamic material: 256–320 kbps, or stay lossless.

One rule matters more than any number: never re-encode a lossy file. Each pass adds a fresh layer of artefacts on top of the last, and quality falls off a cliff after two or three generations. Always transcode from the best source you have.

A Workflow That Saves You Later

Keep one lossless master of anything you would be upset to lose, and treat lossy copies as disposable delivery formats. When you need a smaller version, encode directly from the master rather than from a previously compressed file. Check the result on two very different systems — a decent pair of headphones and a small Bluetooth speaker — because artefacts that vanish on one often shout on the other.

Listen to a difficult passage rather than the loudest chorus. Solo piano, brushed drums and quiet acoustic guitar will tell you far more about whether a bitrate is working than a wall of distorted guitars ever will.

Compression is a trade, not a trick. Understood properly, it lets you carry a library in your pocket with barely a compromise; applied carelessly, it gradually sands the life out of recordings you love. Knowing where that line sits for your own ears is the whole skill — and it is a genuinely enjoyable one to develop.