What sample rate and bit depth should you use?
For most music: 44.1 kHz at 24-bit. For audio going to video: 48 kHz at 24-bit. If your DAW and interface support 32-bit float for recording, use it. That covers maybe 95 percent of people.
The rest of this is the reasoning, and the cases where the answer changes.
Sample rate
Sample rate is how many times per second the audio is sampled. 44.1 kHz means 44,100 samples per second, enough to cleanly reproduce everything humans can hear with margin to spare. It is the CD standard and what streaming platforms deliver.
48 kHz comes from video. Cameras and pro video tools run at 48. If your project is going into video, work at 48 from the start so you do not have to convert at the end.
You can go higher: 88.2, 96, 192. Those have real uses, but also real costs.
Bit depth
Bit depth is the resolution of each sample. More bits means a lower noise floor and more dynamic range to work with.
16-bit is the consumer delivery format. CDs use it. About 96 dB of theoretical dynamic range. Fine for a finished master playing back through speakers, not enough for recording and editing.
24-bit is the working standard. About 144 dB. You can record at conservative levels, leave plenty of headroom, and never worry about losing detail in the quiet parts. There is no reason to record at 16-bit anymore.
32-bit float, where available, goes further. The internal numbers have so much range that the format is effectively unclippable on the way in. You can record a signal well past 0 dBFS, pull it down later, and nothing is lost. A 32-bit float file is about 33 percent larger than a 24-bit one, which is a fair price for the safety net. If your DAW and interface support it, take it.
One more thing worth knowing: most modern DAWs sum in 64-bit float internally, regardless of your file bit depth. The mixing math inside the project happens at that resolution. Your 24-bit files are not the bottleneck.
What I actually use
For my own music I work at 44.1 kHz and record 32-bit float. For mastering, I upsample everything to 96 kHz using a high quality algorithm, and convert to the target rate (usually 44.1 kHz) at the end of the project.
When higher sample rates earn their keep
Nonlinear processing like saturation, distortion and clipping generates harmonics above the original Nyquist frequency. A higher project rate gives those harmonics somewhere to go instead of folding back as aliasing. That is the real argument for 96 kHz in mastering, or in mixes that lean heavily on saturation.
The tradeoffs are real. File sizes double or quadruple. CPU load goes up. Some plugins behave oddly outside 44.1 or 48.
Many plugins do oversampling internally where it matters. That can get you most of the benefit of high rates without committing the whole project to one. But you are then relying on each plugin’s oversampling algorithm, and those vary in quality. If you have several oversampling plugins in a chain, you are also doing repeated rounds of up- and downsampling, which costs CPU and stacks small quality compromises. Past a certain point it makes more sense to work at the higher sample rate so the plugins do not need to oversample at all. That is what I do for mastering. The chain is stacked with nonlinear processing (saturation, clippers, the final limiter), which is exactly where aliasing tends to be audible.
If you want the theory in detail, Monty Montgomery’s piece at Xiph lays out why high sample rates do not help for playback, with math you can actually follow. And Infinite Wave has side-by-side comparisons of how every major DAW handles sample rate conversion. Some are surprisingly bad at it. If yours is bad, render at the working rate and convert with something better at the end.
Mixing rates in one project
What if you are working at 48 kHz and you drop in a sample at 44.1?
Most modern DAWs handle it. Ableton resamples imported audio to the project rate automatically. So does Logic, so do most virtual instrument samplers. You will not notice anything happening, but the resampling can degrade quality depending on the algorithm.
Three things to keep an eye on: what rate your project is set to, what rate your imported files are, and how your DAW handles the mismatch.
I avoid resampling when I can. It is an extra round of conversion you do not need. But do not let it kill your flow. If grabbing a 44.1 loop into your 48 project gets you somewhere creative, do it and worry about the technical purity later.
The thing that actually matters
Sample rate and bit depth are settings. They are not what makes your work sound better. The decisions you make at every step from the arrangement to the mix are.
Mastering is the step after the mix. Another set of trained ears on the finished track, the last check before release. I have been mastering since 2009. How I work.