What Is Lossy Audio?

The permanent trade-off

Lossy compression works by identifying audio data that human hearing is unlikely to notice — and throwing it away. The file gets smaller. The removed data is gone forever. When you play the file back, the decoder reconstructs what it can from what remains.

This is fundamentally different from lossless compression, where data is stored more efficiently but preserved intact. Lossy compression makes a permanent editorial decision: this data probably isn't worth keeping. How much it discards is controlled by the bitrate — lower bitrate means more data removed.

The trade-off is real and worth understanding. At the right bitrate, you'd never know anything was removed. At the wrong bitrate, the missing data becomes audible — and that's when people describe audio as sounding "compressed."

How psychoacoustic compression works

Lossy codecs don't just slice off chunks of audio at random. They use a psychoacoustic model — a mathematical representation of how human hearing actually works — to decide specifically what to remove.

Human hearing has documented limitations. Some sounds genuinely can't be perceived in certain contexts. The codec exploits these limits deliberately. Three mechanisms drive what gets discarded:

Common lossy formats

Does lossy audio actually sound bad?

Not if the bitrate is right. At 192 kbps, most listeners in controlled blind tests (ABX tests, where you hear A and B and try to identify which is the MP3) cannot reliably tell the difference between an MP3 and its lossless source.

The perception of "lossy audio sounds bad" comes from specific situations:

• Very low bitrates (below 96 kbps for music) where artifacts are consistently audible
• Listening on good headphones or reference speakers that reveal compression artifacts
• Complex audio — dense orchestral, acoustic guitar, cymbals — that taxes the psychoacoustic model
• Repeated re-encoding, which compounds the quality loss each time

For podcast voice, background music, casual listening, and streaming: lossy audio at a reasonable bitrate is entirely appropriate. The quality loss is real, but it falls below what most people can perceive in everyday listening conditions.

Generation loss: the real danger

The most significant risk with lossy audio isn't the first encode — it's re-encoding. Every time you convert a lossy file to another lossy format, you're decoding and then re-encoding. The decode step restores slightly degraded audio. The encode step applies another round of psychoacoustic removal to already-damaged audio.

This is generation loss. A single generation of MP3 at 192 kbps sounds fine. After three or four re-encodes — MP3 → AAC → MP3 → OGG — the quality has deteriorated measurably. The artifacts accumulate because the encoder keeps making conservative decisions about what to remove, and each pass removes more.

The practical takeaway: if you're going to distribute audio, do the final lossy encode from a lossless source. Don't repeatedly re-compress the same audio. And never edit or process a lossy file if a lossless original exists — the processing itself can amplify the artifacts.

When lossy is the right choice

Most audio people consume — streaming, podcasts, downloaded music, audio in videos — is lossy. That's not a failing; it's appropriate for the use case.

• Sharing:MP3 or AAC is the right format for sharing audio with others. Small, compatible, good quality.
• Distributing podcasts:MP3 at 128 kbps mono for voice, 192 kbps stereo if you include music. Podcast apps, directories, and listeners all expect MP3.
• Uploading to platforms:YouTube, SoundCloud, Spotify, Apple Music — they all re-encode your upload. Send them the best source you have; they'll do the lossy encode themselves.
• Storage-limited situations:Mobile devices, older hardware with limited space — lossy formats make large libraries practical.

The decision is clear: use lossy for output and distribution. Use lossless for archiving, editing, and production. Keep the lossless master; compress for the audience.