WAV and PCM Audio File Size Calculator

Uncompressed audio is one of the few file formats where size is entirely predictable. There is no compression guessing, no content dependence: a minute of silence takes exactly as much room as a minute of full orchestra.

Three numbers decide everything. Sample rate is how many times per second the signal is measured, so 44,100 Hz means 44,100 snapshots every second. Bit depth is how precisely each snapshot is stored, with 16 bits giving 65,536 possible levels and 24 bits giving over 16,000,000. Channels multiply the whole thing: stereo is two independent streams, 5.1 surround is six.

Multiply them together with duration and you have the answer:

The division by eight converts bits to bytes. At CD settings that comes to 176,400 bytes every second, which is about 10.6 MB per minute.

Bitrate is the same figure expressed per second rather than per file:

CD audio runs at 1,411 kbps. That number is worth remembering because it puts compressed formats in context: a 320 kbps MP3 is roughly a quarter of the data, and a 128 kbps stream about a ninth.

Container overhead is negligible. A WAV header is around 44 bytes, which on a three minute file is far below rounding error.

A handful of setting combinations cover almost all real work, and knowing their per-minute sizes makes storage planning quick.

A three and a half minute song at CD settings comes to about 37 MB. The same performance at 48 kHz and 24 bits is around 60.5 MB, roughly 63% larger for the same music.

The jump from 16 to 24 bits adds half again to the size, because you are storing three bytes per sample instead of two. Doubling the sample rate doubles the size outright. Doing both, going from 44.1 / 16 to 96 / 24, more than triples it.

Long-form recording is where the numbers become real. A 60 minute interview in mono at 48 / 24 is about 518 MB before any editing, and a one hour 5.1 master at the same settings lands near 3.1 GB.

Those are the raw files. Editing multiplies them, because most software writes new audio rather than modifying the original.

Session size scales with track count, not with song length alone. That is the part people underestimate.

Work it out as track count multiplied by recorded duration. A stereo bounce of a four minute song at 48 / 24 is about 69 MB. The 24 track session that produced it, with every track recorded for the full four minutes, is closer to 830 MB in mono tracks. The bounce is the smallest file in the project.

Takes multiply it again. Drum tracking with eight microphones over an afternoon, keeping every take, generates far more audio than the finished song suggests. Comping does not delete the alternatives; it references them.

A rough planning method that works: multiply the expected finished length by the number of tracks, then multiply by three for takes, false starts and safety copies. A five minute song with 16 tracks at 48 / 24 gives 16 × 5 × 8.6 MB, about 690 MB, and 2 GB with the safety factor.

Two habits keep this manageable. Record mono sources as mono files, since a single microphone captured as stereo doubles the size for no information. And decide early whether the project needs hi-res, because converting later does not recover the space already spent.

Back up before you consolidate. Deleting unused audio to reclaim space is the operation most likely to remove something you wanted.

Higher sample rates and bit depths serve production, not delivery. That distinction settles most of the argument.

Bit depth buys headroom. At 24 bits the noise floor sits so far below the signal that you can record conservatively, leaving 18 dB of space above the loudest peak, and still have a clean recording after normalisation. At 16 bits that same caution costs audible noise. This is the strongest practical reason to record at 24 bits regardless of the delivery format.

Sample rate buys processing room. Plugins that generate harmonics, particularly saturation and distortion, alias less at higher rates, and time-stretching has more data to work with. Whether that difference survives into a delivered file is a much weaker claim, since the delivery format itself is usually 44.1 or 48 kHz.

Field recording is the clearest case for going further. 32 bit float capture makes clipping effectively recoverable, which matters when you cannot ask a sound to happen again. The file size cost is worth paying for an unrepeatable take.

Where hi-res is not worth it is delivery to listeners. Streaming platforms transcode everything, and a 96 kHz master is converted before anyone hears it. Uploading the larger file changes nothing except upload time.

FLAC is the sensible middle ground for archives, typically saving 40% to 60% with no loss at all. A 320 kbps MP3 sits near 2.4 MB per minute whatever the source settings were, because the bitrate defines the size.

Confusing kbps with kBps is the classic. Lowercase b is bits, uppercase B is bytes, and the factor of eight between them turns a correct calculation into one that is wildly wrong. CD audio is 1,411 kbps, which is about 176 kBps.

Expecting compressed sizes to follow the source settings is the second. An MP3 at 320 kbps occupies the same space whether it came from a 44.1 kHz or a 96 kHz master, because the bitrate is the setting that determines size. Encoding a hi-res file to MP3 does not produce a larger MP3.

The operating system showing a smaller number than the calculator is not an error. Most systems display binary units while labelling them with decimal names: what they call 1 MB is 1,048,576 bytes, not 1,000,000. The gap is about 4.9% at megabyte scale and grows to 7.4% at gigabyte scale. The bytes are identical; only the division differs.

Recording mono sources into stereo files wastes half the space for nothing. A single microphone has one channel of information, and duplicating it into two does not add any.

Forgetting the multiplier of an edited session catches people out at the worst moment, when a drive fills mid-take. Planning for three times the raw audio is not pessimistic.

One last thing worth separating: file size has no bearing on how loud a master sounds to a listener, since platforms normalise playback. A loudness meter answers that question, and no amount of sample rate will.