MP3, M4A and WAV Sizes: Real Bitrates, Measured

The short answer

Spoken audio inside an MP4 came out as an MP3 of about 40 to 51 MB an hour. Dense stereo came out near 93 MB an hour. Uncompressed WAV is far larger: 11.5 MB a minute for 48 kHz stereo.

How it was measured

Each test file was uploaded through the real converter pages and rendered by REQO’s render farm, which runs ffmpeg 6.1. The output was downloaded and its bitrate taken as bytes times 8, divided by seconds. The measurements were taken on 2026-10-07.

• MP3: LAME at quality 2, a variable bitrate setting, written from the original upload.
• M4A and AAC: ffmpeg’s AAC-LC with a 192 kbps target.
• WAV: 16-bit PCM, so the size is arithmetic and nothing was measured.
• Sample rate and channels: kept as the source has them. Nothing is resampled or mixed down.

The five test sounds are a steady tone, real speech at two sample rates, the same voice on two channels, and a dense synthetic mix made of chords, a kick drum, hi-hat noise and an echo. That last file is not a song. It is sound an encoder has to work as hard for as busy music.

MP3

Stereo speech came out at 114 kbps and the same voice on one channel at 88 kbps. The dense mix needed 207 kbps. A constant 192 kbps file would be 2.2 times the size of the mono speech MP3 for the same words.

That is why a fixed “about 190 kbps” figure does not fit spoken audio. LAME documents roughly that rate for this setting on typical music. In these tests, mono speech came out at about 43% of the dense mix.

M4A and AAC

M4A was asked for 192 kbps. Two channels land on the target (191 kbps and 199 kbps). One channel does not reach it: mono speech came out at 149 kbps. The target is an aim, not a promise, and how far a file falls short depends on the sound. An .aac file is the same encode in a different wrapper and measured 147 kbps on mono speech and 193 kbps on the dense mix.

WAV

WAV here is uncompressed 16-bit audio, so the size is sample rate times channels times 16 bits. These rows are maths, not measurements. They do not move with the sound, which is the opposite of MP3.

Which one to pick

• Spoken recordings you will only listen to: MP3. It is the smallest of the three for a single voice, at 40 MB an hour in these tests.
• Audio you will edit again: WAV. Nothing is lost, and the file is large because of it.
• A file that should keep its quality across phones and players: M4A. Expect it to land near the 192 kbps target on stereo.

To get the audio out of a video, use MP4 to MP3, video to M4A or video to WAV. To turn a WAV into a smaller file, use WAV to MP3.

What this does not cover

• Five test files, one run each. Treat the figures as typical for those shapes of sound, not as limits.
• The music row is a synthetic stand-in. A real song can land above or below it.
• Other encoders and other settings will give other numbers. These are what REQO’s farm writes.
• The WAV rows are calculated, not measured.