Bitrate vs Sample Rate vs Bit Depth: What Each One Does

Your recording app has an export screen with three dropdowns on it. One says 44.1 kHz. One says 16-bit. One says 192 kbps. You have to pick something, and nothing on the screen tells you which of the three is the one that matters.

So you either leave everything on the default and hope, or you turn all three up and end up with a file too big to send.

Bitrate, sample rate and bit depth are three genuinely different measurements, and they do not trade against each other the way that dialog implies. Here is what each one describes, and which one is doing almost all the work on the file you end up with.

Sample Rate: How Often the Sound Got Measured

Sound is a continuous wave of pressure. A computer cannot store continuous anything, so a recorder takes a snapshot of that wave thousands of times per second and writes down a number for each one.

Sample rate is how many snapshots per second. 44,100 Hz, usually written 44.1 kHz, means the wave was measured 44,100 times every second.

That oddly specific figure comes from a rule of sampling: to reconstruct a frequency, you need at least two samples per cycle of it. So a 44.1 kHz recording can represent frequencies up to roughly 22 kHz. Healthy young hearing tops out somewhere around 20 kHz, and falls well below that with age. 44.1 kHz was picked to cover everything a person can actually hear, with a little room above it.

Recording at 96 kHz captures frequencies up to 48 kHz. Bats will appreciate it. Higher rates do have real uses in a studio, mostly to give processing plugins room to work before the final mixdown, but for a voice memo, an interview or a lecture they buy nothing you can hear and double the file.

Bit Depth: How Precisely Each Measurement Was Written Down

Each of those 44,100 snapshots per second has to be stored as a number. Bit depth is how many bits that number gets.

16-bit gives 65,536 possible values per sample. 24-bit gives about 16.7 million.

What that buys is not "detail" in the way people picture it. It is dynamic range: the distance between the quietest sound the file can describe and the loudest, before you run into the noise floor. 16-bit covers roughly 96 dB of range, comfortably more than the gap between a whisper and a shout in any normal room. 24-bit covers far more than a microphone or a room will ever hand you.

So bit depth earns its keep while recording, not while listening. If you record at 24-bit and your interviewee turns out much quieter than you expected, there is enough range underneath to lift the take without dragging hiss up with it. Once the file is mixed and finished, 16-bit describes it perfectly well.

Bitrate: How Much Data the File Spends Per Second

Bitrate is the only one of the three that describes the file rather than the recording. It is simply bits per second of storage.

For an uncompressed file such as a WAV, it is not a choice at all. It falls out of the other two numbers:

44,100 samples/sec × 16 bits/sample × 2 channels = 1,411,200 bits/sec

That is the 1,411 kbps quoted for CD audio. Nothing was decided there. It is arithmetic.

Compressed formats are where bitrate becomes a dial you can turn. Export an MP3 at 192 kbps and you are telling the encoder it gets 192,000 bits to describe each second of sound, and that it must throw away whatever it takes to fit. It discards the parts a psychoacoustic model predicts you will not notice: quiet frequencies masked by louder ones next to them, detail sitting above your hearing.

Give it 320 kbps, the ceiling for MP3, and it discards very little. Give it 64 kbps and it discards a great deal, and you can hear that it did.

The podcaster whose episode won't attach

He records a 50-minute interview as a WAV, because someone told him WAV is the quality option. The file lands at around 500 MB. His co-host's mail server bounces it, the shared drive is full, and the upload to his hosting platform has been sitting at 12% for the last hour. The recording is two people talking in a quiet room. Almost none of that half gigabyte is describing anything either of them said.

The teacher who set every number to maximum

She records her lectures and, wanting the best for her students, sets the recorder to 96 kHz and 24-bit. Her files come out three times the size of the ones her colleague shares. Played back on a laptop speaker, on earbuds, on a phone in a car, they sound exactly like his. She is paying, in gigabytes, for headroom that one voice in a lecture hall was never going to use.

Which Number Actually Moves the File Size

Here is the part the export dialog hides: for a compressed file, bit depth stops existing.

An MP3 has no bit depth. It does not store samples at all. It stores frequency-domain coefficients, and it is measured in bitrate alone. The 24-bit setting you were careful about is gone the moment you encode. Standard MP3 also accepts only a short list of sample rates (32, 44.1 and 48 kHz), so a 96 kHz master gets resampled downward on the way in whether you asked for it or not.

Which leaves a straightforward hierarchy for anything you plan to send to another person:

Number	What it controls	What it does to a finished MP3
Sample rate	The highest frequency captured	Little above 44.1 kHz; higher rates get resampled down
Bit depth	Dynamic range while recording	Nothing. It is not stored in an MP3 at all
Bitrate	Data spent per second	Everything. It sets both the size and the quality

Sample rate and bit depth are recording decisions, made before the microphone is switched on. Bitrate is the delivery decision, made at the end. When a file is too big to send, bitrate is the number to change, and it is very often the only one.

Changing the Numbers on a File You Already Have

The useful thing about that hierarchy is that the delivery decision is the one you can still revisit. You cannot re-record a lecture at a different sample rate, but you can always re-encode what you have at a bitrate that fits.

Both tools below run entirely in your browser, so the audio is never uploaded to a server. That matters more than usual for recordings of other people talking.

How the podcaster gets it sent

He re-encodes the interview at a lower bitrate, picking a value suited to two voices rather than to an orchestra. The projected size updates before he commits, so he can watch the 500 MB WAV land in the tens of megabytes. Same interview, same words, a file his co-host's mail server will accept.

How the teacher gets a lecture she can share

She does not need to re-record anything. She converts the 96 kHz WAV to MP3, which resamples it to 44.1 kHz and drops the bit depth question entirely. Her students get a file that streams on a phone, and it sounds exactly like the version that was three times the size.

The Short Version

Sample rate is how often the sound was measured. Bit depth is how precisely each measurement was stored. Bitrate is how much data the finished file spends per second. The first two are decided before you record, and for speech the defaults are already right. The third decides whether your file sends, and it is the only one worth touching afterwards.

Need a recording to fit? Try Compress audio for free →

Stuck with a WAV or an M4A instead? Audio to MP3 → handles the conversion in the same browser tab.