Manatee

Why converted audio sounds worse than the original

· 6 min read

Audio that sounds worse after a conversion has usually been through a lossy encoder one time too many, at too low a bitrate, or at a sample rate too low to carry the high frequencies. Each leaves a recognizable sound: swishy or watery highs, a muffled top end, a thin and tinny voice. Converting the result again, even “up” to WAV or FLAC, will not bring anything back, so the fix is to go back to the best original you have, keep it as the master, and make every copy from it.

Name the sound

Different faults sound different, which makes a rough diagnosis possible by ear.

  • Swishy, watery or warbling highs. Cymbals that turn to fizz, a sizzle that seems to move around, reverb that bubbles. These are classic low-bitrate artifacts: the encoder ran short of bits and threw detail away in patches that change from moment to moment.
  • Muffled, as if through a wall. The top end is gone. Many encoders cut high frequencies at lower bitrates to save space, or the sample rate was reduced along the way. A file at 22,050 samples per second cannot hold anything above about 11 kHz, and a voice recording at 8,000 samples per second, the rate of phone-style codecs such as AMR, stops at 4 kHz.
  • Thin, tinny or hollow. Common with low-bitrate stereo, where the encoder simplifies the difference between the two channels, and with stereo recordings folded down to mono when the channels partly cancel each other out.
  • A soft whoosh just before drum hits or claps. Called pre-echo. Also a lossy encoder at work, and it gets worse with each further pass.
  • Crackle or harshness on loud passages. Very loud masters can clip after a lossy round trip, because the decoded waveform can overshoot the original’s peaks.
  • Higher and faster, or lower and slower. That is not a quality problem but a sample-rate mix-up, covered in an audio file that plays too fast or sounds high-pitched.

The usual causes

A second lossy pass. MP3 to M4A, AAC to MP3, a voice memo turned into an MP3 to send. The second encoder cannot tell the first one’s damage from the music, so it spends bits on the artifacts and discards more of what was left. One pass at a good bitrate is usually inaudible; two or three passes at modest bitrates usually are not. Lossless vs lossy explains the mechanics. The practical habit is to count how many lossy steps a file has already been through, including ones you did not do yourself: a download from a streaming or video site, or a messaging app that re-encoded a voice note before it reached you.

A low bitrate. A 64 or 96 kbps MP3 is fine for speech and noticeably worse for music. If a converter chose a low bitrate to keep the file compact, the result sounds like it.

A low-quality source. Converting a 96 kbps MP3 to a 320 kbps MP3 does not lift it to 320. It stores the 96 kbps sound, artifacts and all, in more bits, and adds a generation of loss on top.

Sample-rate conversion. Music is usually 44.1 kHz and audio from video is usually 48 kHz, so conversions between the two happen constantly. A good converter does this without audible change. Repeated or careless conversions can add a faint harsh or metallic edge.

Converting up does not undo it

Turning an MP3 into WAV or FLAC makes a much larger file with exactly the same sound. In our testing on macOS 15, a 96 kbps MP3 converted to WAV with Apple’s own afconvert command came out about 15 times larger and reported a bit rate of 1,411,200 bits per second, the figure for CD audio. The number says CD; the sound is still the 96 kbps MP3. A large WAV is not evidence of a good source.

It also sets a trap. If that WAV replaces the MP3 in your files, the next time someone converts it back to MP3 the audio goes through another lossy pass, and nobody can tell from the file that it already had one.

Check what a file really is

  1. Finder. Select the file, press Command-I and expand More Info. For most audio formats it shows the sample rate and the bit rate.
  2. Terminal. The built-in afinfo command prints the format, channels, sample rate and bit rate:
afinfo ~/Music/interview.mp3

For the 96 kbps test file above, the useful lines read:

Data format:     2 ch,  44100 Hz, .mp3 (0x00000000) 0 bits/channel, 0 bytes/packet, 1152 frames/packet, 0 bytes/frame
bit rate: 96000 bits per second

For a lossy file, the bit rate tells you roughly how much was kept. For a WAV or FLAC, the numbers only describe the container. They cannot tell you whether the audio inside was once an MP3.

  1. Compare fairly. Play both files in QuickTime Player, at the same volume, switching at the same passage. A version that is even slightly louder tends to sound better, and the Music app can apply its equalizer or Sound Check to library tracks, so listen to both in the same place with the same settings.

What to keep as the master

  • Keep the first-generation file. The original recording, the lossless rip, the download in the best format you were offered. If all you ever had was an MP3, that MP3 is your master: keep it unchanged and never replace it with a converted copy.
  • Make every copy from the master, not from another copy. Need an M4A for your phone and an MP3 for a client? Make two conversions from the master, not one from the other.
  • Record lossless if you will edit or convert later. Voice Memos on the Mac has an Audio Quality setting in its Settings with Compressed and Lossless choices. Lossless files are far larger, and they give you a clean master to convert from.
  • Choose the bitrate on purpose. For music you care about, 256 kbps AAC or 320 kbps MP3 for the copy you send; for speech, lower is fine.

Manatee converts audio on your Mac. It accepts 122 input types, including FLAC, OGG, Opus, WMA and AMR, and writes MP3, M4A, WAV or FLAC. The converted file goes next to the source, which is never changed, so the master stays the master. It cannot restore what an earlier lossy pass removed, and nothing can: a FLAC made from an MP3 sounds like the MP3. After converting, check the new file with afinfo or Get Info so you know exactly what you made.

Questions

Is M4A better than MP3? At the same bitrate, AAC in an M4A usually sounds better than MP3. Converting an existing MP3 to M4A, though, adds a generation of loss and makes it slightly worse, not better.

Why does the converted file sound fine on my phone but bad on good speakers? Phone speakers do not reproduce much of the high-frequency and stereo detail that low-bitrate encoding throws away. Good headphones or speakers reveal what is missing.

Can software put the missing detail back? Some tools generate a plausible top end, but that is invention, not recovery. If a better original exists anywhere, such as the sender’s file, the source recording or a lossless purchase, getting it is worth far more than any processing.