Sample Rate & Bit Depth Explained

The two settings answer different questions. Sample rate decides which frequencies can be stored; bit depth decides how quiet a detail can be before it disappears into the noise floor. The values below are the ones you will meet when exporting music.

You deliver songs to Spotify and Apple Music through a distributor, so the distributor's upload rules apply first; the table shows what the platforms themselves publish. Spotify's instruction is short: deliver the audio exactly as it was mastered, keep the original sample rate and bit depth, and let Spotify do any conversion. Apple asks for the same thing in different words: send the highest native resolution you have, and do not upsample or bit-pad a 16-bit file to 24-bit. The YouTube row covers videos you upload yourself, such as a visualizer or lyric video. A release sent through a distributor reaches streaming services from the master you gave the distributor, not from a video file.

Digital audio stores a sound wave as a series of measurements. The sample rate is how many measurements are taken per second, in hertz: at 44,100 Hz the wave is measured 44,100 times every second. The Nyquist-Shannon sampling theorem says a signal can be reconstructed exactly if it is sampled at more than twice its highest frequency. So a 44.1 kHz file can hold frequencies up to 22.05 kHz, slightly above the upper limit of human hearing (about 20 kHz, and lower for most adults). Before sampling, an anti-aliasing filter removes content above half the sample rate. Without it, those frequencies would fold back into the audible range as aliasing, a distortion unrelated to the original sound. The gap between 20 kHz and 22.05 kHz gives that filter room to work.

Work at the rate your project needs and keep it through to the master. For music that will be released as audio, 44.1 kHz or higher is fine; if you already recorded and mixed at 48 kHz or 96 kHz, deliver that, because Spotify and Apple both ask for the native rate rather than a conversion. For audio that will sit under video, use 48 kHz. The two rates have different histories. 44.1 kHz comes from early digital recording on video tape: PCM adaptors stored 3 samples per video line, and 3 × 245 lines × 60 fields (NTSC) and 3 × 294 lines × 50 fields (PAL) both give 44,100 samples per second. The CD standard kept that rate. 48 kHz became the video and film standard because it divides evenly into common frame rates: 2,000 samples per frame at 24 fps and 1,600 at 30 fps. That makes audio and picture easy to keep in sync, which is why YouTube recommends 48 kHz for uploads. Converting between the two with a good sample-rate converter is not audible in practice, but every conversion is a processing step. Do it once, at the end, and only when the destination needs it.

Bit depth is the number of possible values each sample can take: 2^16 = 65,536 for 16-bit and 2^24 = 16,777,216 for 24-bit. More values mean smaller rounding errors, which sit as a noise floor under the music. Each bit adds about 6 dB of dynamic range, so 16-bit reaches about 96 dB and 24-bit about 144 dB. In practice, converters and rooms are noisier than 24-bit allows, so 24-bit is effectively noise-free headroom: you can record at sensible levels without the quiet parts approaching the noise floor. If you need a 16-bit file, for a CD or a service that only takes 16-bit, add dither when you reduce the bit depth. Dither is a very low level of noise that turns rounding distortion into a smooth, benign hiss. Reduce bit depth once, as the last step; never go back up by padding a 16-bit file to 24-bit, because the extra bits contain nothing.

A 32-bit floating-point sample stores a value with 24 bits of precision plus an exponent, so precision stays the same whether the signal is quiet or loud. The practical benefit is that levels above 0 dBFS inside a DAW or a 32-bit float file do not clip; you can turn them down later and the peaks come back intact. That makes 32-bit float a good working format for recording and mixing. It is not what most distributors expect for delivery, and Spotify reduces anything above 24-bit internally. Bounce your final master to 24-bit (with the true peak below your ceiling) unless your distributor documents something else.

High-resolution audio means sample rates above 48 kHz and/or bit depths above 16 bits, usually 24-bit at 96 or 192 kHz. Apple, for example, calls audio up to 48 kHz lossless and audio above 48 kHz, up to 192 kHz, hi-resolution lossless. Whether listeners can hear the difference is contested. A 2016 meta-analysis of eighteen published experiments (over 400 participants and 12,500 trials) by Joshua Reiss, published in the Journal of the Audio Engineering Society, found a small but statistically significant ability to tell high-resolution audio from 16-bit, 44.1 or 48 kHz audio, and a much larger effect when listeners had been trained. It did not find a large, obvious difference for untrained listeners. The clearer benefits are in production: higher rates move the anti-aliasing filter further above the audible range and give pitch-shifting and time-stretching more material to work with. For delivery, the platforms' advice is the useful rule: send what you have natively, and do not upsample.

Your DAW or audio editor shows both values in its file or project information, and the command-line tool ffprobe (part of FFmpeg) prints them: run ffprobe -hide_banner yourfile.wav and read the stream line, which shows the codec, sample rate and sample format (for example pcm_s24le, 48000 Hz, stereo). The free check on this page reads the same information from the file's header, in your browser; the audio is not uploaded. It reads the header rather than decoding the file because browsers decode audio at the playback device's sample rate, as 32-bit floating point, so a decoded file no longer tells you what the original was.