Why converting audio can make it worse
Every pass through a lossy encoder discards a little more, and it never comes back. Knowing when that matters — and when it genuinely does not — saves both quality and a lot of pointless converting.
Last reviewed
What lossy compression actually removes
MP3, AAC, Vorbis and Opus all work the same way in principle: they model what human hearing can and cannot detect, and store only what you would notice. A quiet sound immediately after a loud one is masked and can be discarded. Frequencies above what the material contains need no bits. Stereo information that is nearly identical in both channels can be stored once.
The result is astonishing compression — a CD track at around a tenth of its size with no audible difference — and it is not reversible. The discarded information is gone from the file, and no decoder, converter or enhancement recovers it. That is the whole meaning of "lossy", and everything below follows from it.
Generation loss: why a second pass hurts more than the first
Here is the part that surprises people. Encoding a clean recording to MP3 at a decent bitrate is nearly transparent. Taking that MP3, decoding it and encoding it again — even at the same bitrate, even with the same encoder — is audibly worse, and the reason is specific.
The second encoder cannot tell the first encoder's artefacts from real sound. What it receives is not the original recording; it is the original minus what was removed, plus the small distortions the removal introduced. Those distortions are, to a psychoacoustic model, ordinary audio content — so the encoder faithfully spends bits preserving them, while throwing away a further slice of what is left.
It is worse still between different codecs, because they disagree about what is inaudible. Converting AAC to MP3, or Opus to MP3, means the second encoder is preserving decisions the first one made for reasons it does not share. This is why a transcoded file can sound worse than either codec would at half the bitrate from a clean source.
The one rule
Keep a lossless original, and generate every delivery copy from it.
That is the whole discipline. If you have a FLAC or a WAV, every MP3 you make from it is a single clean generation — and if you later want a different bitrate, a different format, or whatever replaces MP3 next, you convert again from the original and lose nothing extra. Converting FLAC to MP3 is the best case any lossy encoder ever gets, precisely because there is no earlier damage to preserve.
Delete the lossless original and every future conversion starts from something already degraded. That is a decision with a long tail, and it is usually made to save disk space that costs less than the music did.
When it genuinely does not matter
Being precise about this is as useful as the warning, because a lot of advice in this area is superstition.
One transcode at a sensible bitrate is usually inaudible. Converting a 192 kbps AAC to a 256 kbps MP3 to play in an old car stereo is fine. People hear about generation loss and conclude that any conversion ruins a file; it does not, provided you give the second encoder room.
Encode a little above the source. If the original is 128 kbps, encoding to 128 compounds two lots of loss at the same budget; encoding to 192 gives the second pass enough bits that it adds nothing you can hear. Going to 320 does not make it better than the source — you cannot exceed what you started with — but it stops the second pass from taking anything.
Speech is far more forgiving than music. A lecture or a podcast at 96 kbps in mono is perfectly clear, and transcoding it barely registers. Cymbals, applause and dense reverb are where low bitrates and second generations become obvious.
What WAV does and does not do
Converting a lossy file to WAV is worth understanding because people reach for it expecting the wrong thing. Turning an MP3 into a WAV stores the decoded audio uncompressed, which stops any further loss and does not undo the first one. The WAV is ten times the size and contains exactly the sound the MP3 contained.
That is still the right move before editing. Cutting, mixing and adjusting a file means saving it repeatedly, and every save of a lossy file is another generation. Convert to WAV, do all the work, encode once at the end — one generation of loss instead of five.
One practical limit: a WAV file cannot exceed 4 GB in the common format, which a long recording at high resolution can reach. Long-form work belongs in a format built for it.
Frequently asked questions
Can I improve an MP3 by converting it to a higher bitrate?
No. The file gets larger and contains the same sound plus the new encoder's guesses. Nothing can recover information that was discarded — an upscaled MP3 is a bigger file, not a better one.
How many times can I convert before I hear it?
On music at moderate bitrates, a second generation is often audible on difficult material — cymbals, applause, dense mixes — and a third usually is. On speech you can get away with several. The honest answer is that it depends on the material, which is why the rule is to avoid the situation rather than to count passes.
Is MP3 still the right choice in 2026?
Only for compatibility, which is a real reason. AAC beats it at every bitrate and Opus beats both, and every current phone and computer plays them. Choose MP3 when something older in the chain requires it, not by default.
Does extracting audio from a video re-encode it?
It depends on what you ask for. Copying the existing track out is lossless and instant, because nothing is decoded. Asking for MP3 or WAV specifically means a conversion — so if you just want the audio, copying the original track is both faster and better.
Why does my 320 kbps MP3 sound worse than a 128 kbps original?
Almost certainly because it was transcoded from something already lossy. Bitrate tells you how much room the last encoder had, not how good the source was — a 320 kbps file made from a 96 kbps original carries all of that original's damage in a much larger file.