What Sample Rate and Bit Depth Should I Record At?

Record at 48 kHz, 24-bit. The 24 bits give you about 144 dB of theoretical dynamic range against 96 dB for 16-bit, so you can track with plenty of headroom and lose nothing. The 48 kHz captures everything up to 24 kHz, past the edge of human hearing, and it is the native rate for video, broadcast and Dolby Atmos. Going to 96 kHz doubles every file, a 40-track, four-minute song goes from 1.38 GB to 2.76 GB, to capture frequencies nobody in the room can hear.

I get this question a lot, often from someone who has already recorded half an album and is worried they picked wrong. Most of the time they did not.

I have mixed and mastered for 20 years, 19 Billboard Top 20 albums, records past 3.3 billion streams, and I am on Apple's Digital Masters list, so I see what arrives in the inbox and what the platforms ask for on the way out. Here is the math behind the answer.

Sample Rate: How High Can the Recording Go

Sample rate is how many times per second the converter measures the signal. The highest frequency a rate can capture is half of it, the Nyquist limit. So 44.1 kHz captures up to 22.05 kHz, 48 kHz up to 24 kHz, 96 kHz up to 48 kHz. Adult hearing tops out around 20 kHz, and for most of us it is lower than that by the time we are paying for studio time.

Every doubling of the rate doubles the file sizes and roughly doubles the work your computer does per plugin. Here is what that looks like, computed for 24-bit mono tracks:

Sizes are uncompressed WAV: sample rate × 3 bytes (24-bit) × 60 seconds. Latency is one buffer, one direction; your round trip through an interface will be longer.

The step from 44.1 to 48 costs you almost nothing, about 9% more disk. The step from 48 to 96 costs you double everything, and what you buy is content between 24 kHz and 48 kHz.

When a Higher Rate Actually Earns Its Keep

I am not going to pretend 96 kHz does nothing. It does three real things, they are just narrower than the marketing.

• Lower latency at the same buffer. If you are tracking through plugins and the singer hears the delay in their headphones, doubling the rate halves the buffer time. At 128 samples that is 2.67 ms down to 1.33 ms. Dropping the buffer size at 48 kHz gets you most of the same thing.
• Less aliasing from some plugins. Saturation and clipping generate new high harmonics, and a plugin that does not oversample internally can fold those back down into the audible range. At 96 kHz there is more room above the band before they fold. Most modern saturation plugins oversample on their own, so this matters less every year.
• Extreme pitch and time work. If you are going to slow a sound down two or three octaves for sound design, the content that was above 20 kHz becomes content you can hear. That is a real reason, and it is a sound-design reason, not a song-recording one.

Outside those three, the higher rate is disk and CPU spent on nothing you can hear.

Bit Depth: How Much Room Before the Noise

Bit depth is how finely each of those samples is measured, and it sets the distance between the loudest thing the file can hold and its own noise floor. The math is about 6 dB per bit: 16-bit gives about 96 dB, 24-bit about 144 dB.

No converter gets near 144 dB, and no room is quiet enough to need it. That is the point: at 24-bit you can track with peaks well below zero, a singer who leans into the chorus does not clip, and the quiet parts stay nowhere near the noise.

So there is no decision to make here for tracking. 24-bit, every time.

32-bit float is the one that confuses people. Your DAW already mixes internally in 32-bit float, which is why a channel can go over zero inside the session and come back down without damage. Saving your recordings as 32-bit float files is a different claim. On a normal interface the converter is 24-bit, and if the signal clips the converter, the take is clipped regardless of how the file is saved. The exception is the handful of field recorders built with dual converters specifically for float recording. For a home or studio session through an interface, 24-bit files are the right call.

Where 32-bit float does matter is delivery. What I deliver masters as, for what it's worth, is 48 kHz, 32-bit WAV.

What the Platforms Actually Ask For

None of the platforms require you to record at a high rate. What they care about is that the file you hand them has not been degraded on the way:

• Spotify wants a lossless file, at least 44.1 kHz, and 24-bit if that is what your master is. Anything below 16-bit / 44.1 kHz is not eligible for its lossless tier. How to put a song on Spotify has the full upload checklist.
• Apple Digital Masters needs a 24-bit source, 44.1 kHz minimum, at its native rate, never upsampled or bit-padded. That last part is the one people break. The Apple Digital Masters requirements walks the whole spec.
• Dolby Atmos music work is built around 48 kHz, which is one of the quieter reasons 48 has become the default: if there is any chance the record gets a spatial mix, you are already there. How to release music in Dolby Atmos covers the rest.

Notice what "never upsampled" means for you. Converting a 44.1 kHz session to 96 kHz does not add anything that was not captured. It just makes a bigger file of the same recording, and a certified engineer is not allowed to pass it off as high resolution.

The Mistakes I Actually See

Keeping the rate consistent matters more than which rate. What actually costs people time:

• Mixed rates in one session. Drums tracked at a studio at 48, vocals done at home at 44.1, imported without converting. Depending on the DAW the vocal plays back at the wrong speed and pitch, or gets converted on import without anyone noticing.
• Bouncing at a different rate than the session. A 48 kHz session exported at 44.1 adds a sample rate conversion before I have touched a fader. Export at the session rate and let the mastering stage do any conversion once, at the end. How to prepare tracks for mixing has the export settings I ask every client for.
• Recording too hot because the meter "should" be near zero. That rule is a leftover from 16-bit tape-era thinking. At 24-bit, peaks around -10 dBFS are plenty. You give up nothing and you never lose a take to a clip.

The Short Version

Set the session to 48 kHz, 24-bit before you record the first note, and never change it for the life of the project. Track with your peaks comfortably below zero. Export stems at the session rate, and let the master be where any format decisions get made. If you have been recording at 44.1 all along, keep going at 44.1; switching mid-album is worse than either number.

If you are recording at home and want the vocal side of this in detail, how to record vocals at home picks up where this leaves off. And when the files are ready for someone else's ears, here is what mixing and mastering cost, itemized honestly, alongside what a producer costs if the project needs more than a mix. Or skip the ranges and get a number calibrated to where you actually are as an artist from the rate calculator.