The one line of arithmetic
Uncompressed audio is bytes equals sample rate times bit depth divided by eight, times channels, times seconds. There is nothing else in it. One minute of 44.1 kHz 16-bit mono is 44100 x 2 x 60, which is 5,292,000 bytes, about 5.3 MB. One minute of 48 kHz 24-bit stereo is 48000 x 3 x 2 x 60, which is 17,280,000 bytes, about 17.3 MB, or 16.5 MiB as an operating system will report it.
Everything that makes real sessions bigger than expected is a multiplier on that line, not a change to it. Track count multiplies it. Rolling time multiplies it, and rolling time is far longer than song length because tracks keep writing while nobody is playing. Takes and playlists multiply it again. Bounces and stems add a fixed amount per export. And then the whole thing gets multiplied by however many places you keep it.
GB and GiB, and why the drive looks smaller than the box said
Storage is sold in decimal gigabytes, where a GB is 1,000,000,000 bytes. Most operating systems report in binary units, where the equivalent unit is 1,073,741,824 bytes, properly called a gibibyte. The ratio is about 7.4 percent at this scale, which is why a drive labelled 1000 GB shows up as roughly 931 of the units your file manager calls GB.
This page shows both because the two numbers are used for different things. Compare against the label on a drive using the decimal figure. Compare against what your system says is free using the binary one. Neither is wrong and neither is a rounding error.
Sample rate, bit depth and the headroom conversation
The two dimensions do completely different jobs and it is worth separating them, because the storage cost is identical per doubling and the benefit is not.
Sample rate sets how high a frequency can be represented, at half the sample rate. 44.1 kHz covers everything a human ear responds to, and going higher is about giving anti-alias filters more room to work and about how some non-linear processing behaves, not about hearing more of the recording. Doubling the rate doubles the file, doubles the disk throughput and roughly doubles the processing load, and the last of those usually becomes the limit first.
Bit depth sets the distance between the loudest signal that fits and the noise floor underneath it. Sixteen bits gives about 96 dB of that range; twenty-four gives about 144, which is far more than any microphone and preamp will deliver. That surplus is the point. It means you can set the recording level conservatively, peaking well below the ceiling, and still have a noise floor buried below anything the room contributes. Recording quiet at 24-bit costs nothing, and it removes the risk that the one take that mattered clipped on a moment nobody predicted.
That is the headroom conversation in one sentence: with 24-bit you do not need to chase the top of the meter, so do not. The old habit of tracking as hot as possible came from 16-bit tape-era thinking where the noise floor was a real problem. It is not a real problem now, and hot tracking costs you the safety margin instead.
32-bit float is a different proposition again. It stores a value and an exponent rather than a fixed-range integer, so it has no hard ceiling and a take recorded far too loud can be brought back down without damage. It is half again the size of 24-bit and it buys insurance, not resolution. Whether that trade is worth it depends on whether your sessions are ones where you get a second attempt.
Where the space actually goes
| Cause | Typical effect on session size |
|---|---|
| Rolling between takes | Often doubles the recorded minutes against the music |
| Playlists and alternate takes | 1.3 to 2 times, higher on comped vocals |
| Stereo rather than mono tracks | 2 times on those tracks |
| 48 kHz to 96 kHz | 2 times everything |
| 24-bit to 32-bit float | 1.33 times everything |
| Bounces, stems and rough mixes kept | Each one costs a stereo file of the full length |
| Backups | Multiplies the final figure by the number of copies |
The item people forget is the first one. A three-hour session does not contain three hours of music, but it does contain three hours of every armed track, and the disk does not know the difference between a take and a conversation.
Backups are part of the size, not an afterthought
A session that exists in one place is not stored, it is at risk. The usual working rule is three copies on two different kinds of media with one of them somewhere else, and whatever version of that you adopt, the honest storage figure for a session is the session size times the number of copies. Budget for it up front, because discovering it after a session is how projects end up with one copy and a good intention.
The practical corollary is that format choices are multiplied by the copy count too. Tracking at 96 kHz 32-bit float and keeping three copies is four times the storage commitment of 48 kHz 24-bit with the same three copies, for the same performance.
Where this connects
For a delivery file rather than a session, size comes from bitrate rather than from sample rate and bit depth, which is what the media file size calculator covers, and the video version of the same question is on the video file size calculator. For the timing side of the session, the BPM delay time calculator converts tempo to milliseconds and the metronome plays it. The room those recordings are judged in is covered by the acoustic treatment calculator.
Questions people ask
How do I calculate the size of a WAV file?
Bytes equals sample rate times bit depth divided by eight, times the number of channels, times the length in seconds. A 48 kHz 24-bit stereo file running four minutes is 48000 x 3 x 2 x 240, which is 69,120,000 bytes, about 69 MB. Add a small fixed amount for the header, which is trivial at any useful length. The formula is the same for AIFF and for any other uncompressed linear PCM format, because they all store the same samples and differ only in how the file is wrapped.
Is 24-bit worth the extra space over 16-bit?
For recording, yes, and the reason is operational rather than audible. Twenty-four bits gives roughly 144 dB between the ceiling and the noise floor, which is more than any microphone and preamp chain can produce. That surplus lets you set levels conservatively, peaking well below the top, without the noise floor becoming audible. You get a large safety margin against unexpected loud moments for a 50 percent increase in file size. For delivery, 16-bit with proper dither is entirely sufficient and is what most distribution formats want anyway.
Should I record at 96 kHz?
It depends on what you are doing, and the honest answer for most sessions is that the benefit is small and the costs are not. Higher rates give anti-alias filters more room and can change how some non-linear processing behaves, which matters more on heavily processed electronic material than on a straightforward acoustic capture. What you pay is double the storage, double the disk throughput and roughly double the plugin load, and it is usually the processing that runs out first. If you are unsure, 48 kHz 24-bit is the working default for a reason. Decide it before the session, because changing rate mid-project causes more trouble than it saves.
Why does my drive show less free space than the label says?
Two separate reasons stacked. First, drives are sold in decimal gigabytes of 1,000,000,000 bytes while most operating systems display binary units of 1,073,741,824, which accounts for about 7.4 percent at this scale. Second, the file system itself takes space for its own structures, and some systems reserve a further slice. A drive labelled 1000 GB commonly presents around 931 of the units your file manager calls GB, and a little less again after formatting. Neither is a fault, and this calculator shows both conventions so you can compare against whichever number you are looking at.
How much space should I keep free while recording?
Enough that a session running long does not become a crisis, which in practice means more than one session of headroom and preferably several. A drive close to full also behaves worse: writes get slower and more fragmented, which on a recording drive turns into dropouts rather than a mild delay. The other habit worth having is to check the figure before people arrive rather than during a take. Running out of space mid-performance is the one storage failure that costs something you cannot recreate.