MP3, AAC and WAV: choosing an audio format and bitrate

Compare the main audio formats, understand bitrate and learn which settings suit speech, music and archives.

Audio files come in a confusing range of formats, yet in day-to-day work only a handful matter. Choosing well is mostly a matter of answering three questions: where will the file be played, how much space is available, and will you edit it again later?

How digital audio is stored

A microphone produces a continuous signal, and a recording samples it many times per second. CD audio takes 44,100 samples per second at 16 bits per sample for each of two channels. Multiply those together and you get 1,411 kilobits per second, or roughly 10 MB per minute. Uncompressed audio is simple and exact, but it takes a lot of space.

The sample rate limits the highest frequency that can be represented. Rates of 44.1 kHz and 48 kHz cover the range of human hearing. Music generally uses 44.1 kHz, and video production generally uses 48 kHz. Bit depth determines how much dynamic range the recording can hold, and 24-bit files are common during recording since they leave more room for mistakes.

Uncompressed: WAV

WAV normally contains uncompressed audio, and it is the standard working format in editing and production. It suffers no generation loss, opens in nearly all audio software and requires very little processing power. Its weakness is size, which makes it poor for sharing and for phones with limited storage.

Lossless compression: FLAC and ALAC

FLAC reduces file size, often to about half or two-thirds of WAV, and decompresses to precisely the original samples. It is a sensible choice for a music archive. Apple's equivalent is ALAC. Lossless files can be converted to any other format later without first suffering a lossy step.

Lossy compression: MP3, AAC and Opus

Lossy formats remove parts of the sound that are masked by other sounds and therefore harder to hear. MP3 is the oldest and has near-universal support. AAC usually sounds a little better than MP3 at the same bitrate and is used widely on Apple devices and streaming services. Opus performs very well, especially for speech and at low bitrates, and is common in calls and web applications, though support in older hardware is more limited.

Which bitrate should you choose?

Bitrate for lossy audio is a compromise between size and quality, and these are reasonable guidelines:

  • Speech, podcasts and lectures: 64-96 kbps mono is often enough, or up to 128 kbps stereo.
  • Everyday music listening: 128-192 kbps AAC or 192 kbps MP3.
  • Music where quality matters: 256-320 kbps.
  • Editing or archiving: WAV or FLAC instead of a lossy format.

To estimate size, multiply kilobits per second by the seconds of duration and divide by 8. A 60-minute podcast at 96 kbps is 96 × 3600 ÷ 8, or 43,200 KB: about 43 MB.

Going above 192-256 kbps yields a diminishing return that many listeners cannot hear on ordinary headphones. A short listening test of your own material, without knowing which file is which, is worth more than any general rule.

Mono, stereo and speech recordings

Stereo stores two channels. A voice recorded with a single microphone is effectively mono, so storing it as stereo wastes data. Converting speech recordings to mono gives you a better result at the same bitrate. Music with real stereo placement should remain stereo.

Extracting audio from video

Video files usually carry an AAC or similar track. Converting that track to MP3 means decoding one lossy format and encoding another, so a small additional loss occurs. If the destination accepts AAC, extracting it into an AAC file avoids unnecessary loss. Choosing a higher MP3 bitrate than the source does not bring quality back, as a 128 kbps source remains a 128 kbps source no matter how it is saved.

WAV extraction is useful when you plan to edit the audio. Do your editing there and create a compressed copy at the very end, so you avoid stacking several generations of lossy compression.

Loudness and clipping

If a track sounds quiet compared with another, simply raising its volume can push the peaks past the digital maximum, causing crackling distortion known as clipping. Normalising to a target loudness is safer than raising the level by a large amount. Streaming and podcast platforms each apply their own loudness targets, so an extremely loud master has little to gain.

Respecting rights

Having a file does not mean being free to reuse or publish its contents. Convert recordings you made yourself, have a licence for or are otherwise allowed to use, particularly when a recording will be republished or used in a commercial project.

Summary

  • Sending audio to anyone, anywhere: MP3 at 192 kbps.
  • Listening on modern devices with less space: AAC.
  • Editing and production: WAV.
  • Long-term music archive: FLAC.
  • Voice-only recordings: mono at a lower bitrate.

Tools mentioned in this guide