Every WAV file is a list of numbers. Two settings decide what those numbers are: how many of them there are per second, and how finely each one is measured. They are called sample rate and bit depth, and an export dialog asks for both every time you bounce a loop.
This guide explains what the two settings mean, which values producers actually use, how big the file is, and what to pick when the loop is going to another producer rather than a listener.
What sample rate and bit depth mean
Digital audio measures the sound wave many times a second. Each measurement is a sample.
- Sample rate is how many samples are taken per second, in hertz. 44,100 samples per second is written 44.1 kHz.
- Bit depth is how many binary digits each sample has, which is how many distinct levels it can land on. 16 bits gives 65,536 levels. 24 bits gives about 16.8 million.
Neither setting changes how long the loop is or what notes are in it. They change how precisely the wave is recorded.
Sample rate: 44.1 or 48 kHz
The rule behind it is the Nyquist limit: a sample rate can capture frequencies up to half its value. At 44.1 kHz that is 22.05 kHz, and at 48 kHz it is 24 kHz. Human hearing is usually quoted as ending somewhere around 20 kHz, so both rates cover everything a listener can hear. That is why 44.1 kHz was chosen for CDs, and why 48 kHz became the standard for video and film.
So the practical difference between them is small. 48 kHz is not "better" in a way anyone can hear on a finished loop; it is simply the rate that video work runs at. 44.1 kHz is the common default for music. Higher rates such as 88.2 or 96 kHz exist, and some producers like them for heavy pitch-shifting or time-stretching, but they make every file bigger and a loop bound for a stranger's session gains almost nothing from them.
Bit depth: 16, 24 or 32-bit float
Each extra bit roughly doubles the number of levels, which adds about 6 dB of dynamic range. That gives the two numbers you will see quoted: about 96 dB for 16-bit and about 144 dB for 24-bit.
What that range buys you is room between the loudest sound and the noise floor. At 16-bit, a quiet part of the loop sits closer to the hiss of rounding error, and if a collaborator turns it up by 20 dB, the noise comes up with it. At 24-bit the floor is so far down that you can record, export and then raise a quiet signal without hearing it. 24-bit is the everyday choice for working files for exactly that reason.
32-bit float stores each sample as a number with an exponent, rather than as a fixed level. It can hold levels far above 0 dBFS without clipping inside the file, and far below the floor of any real converter. That makes it useful as a safe intermediate format: a stem that was exported too hot will not be permanently flattened in the file. It makes files a third bigger than 24-bit. If you are not sure, 24-bit is enough.
Why the session rate matters to a collaborator
When another producer drops your loop into a project running at a different rate, their DAW has to convert it. Most DAWs do this on import or on the fly, and a good sample rate converter is transparent. But it is one more thing that can go wrong: a project that has been silently playing a 48 kHz file at 44.1 kHz runs about 8 percent slow and about a semitone and a half flat, and an unconverted file can drift out of time against everything around it.
You cannot know what rate a stranger works in, so the answer is not to guess. Export at the rate you actually worked at, do not convert upwards to look professional, and label it. Resampling a 44.1 kHz loop to 96 kHz adds file size and no information. The detail above 22 kHz was never recorded, and a converter cannot invent it.
ColabLoop takes this approach as well. A loop can be uploaded at 44.1 or 48 kHz, the upload reads and stores the actual rate, and the loop's page shows it, so a collaborator knows what they are opening before they download. The upload form accepts any uncompressed PCM WAV, which covers 16-bit, 24-bit and 32-bit float.
How big is a WAV file?
A WAV is uncompressed, so its size is plain arithmetic:
bytes per second = sample rate x bytes per sample x channels
where bytes per sample is the bit depth divided by 8. A WAV adds a header of 44 bytes on top, which is too small to matter for anything longer than a moment.
For a 30-second stereo loop at 48 kHz and 24-bit:
- 48,000 samples x 3 bytes x 2 channels = 288,000 bytes per second
- x 30 seconds = 8,640,000 bytes, plus the 44-byte header = 8,640,044 bytes, about 8.2 MB
The same loop at 44.1 kHz is 264,600 bytes per second, so 7.9 million bytes. In 16-bit at 44.1 kHz it is 176,400 bytes per second, or 5.3 million. In 32-bit float at 48 kHz it is 384,000 bytes per second, or 11.5 million. Mono is exactly half of stereo.
On ColabLoop a loop must be between 10 and 60 seconds long and no more than 30 MB. At 48 kHz, 24-bit stereo, a full 60-second loop is 17.3 million bytes, so the 60-second limit always binds before the size limit does. You would need a 96 kHz, 24-bit stereo file to find the size cap, and even then it only starts to matter at about 54.6 seconds. For the usual four- or eight-bar loop, how long a loop should be has a table worked out for your tempo.
Mono or stereo
Stereo carries two channels, so it is twice the size of mono at the same rate and bit depth. That cost is worth paying if the loop has any width in it: panned guitars, a stereo pad, a wide reverb.
Mono is the better choice for a part that is mono by nature, such as a single bass line or a vocal. A stereo file whose two channels are identical wastes half its size and implies a width that is not there. If a loop has both, a mono sub bass and a stereo top, export it stereo and keep the low end centred. That is how most producers expect to receive it.
Dither when you reduce bit depth
Dither only matters when you go down in bit depth, for example from a 24-bit project to a 16-bit file. Cutting the extra bits off leaves a rounding error that is correlated with the signal, and on quiet material it sounds like a thin, gritty distortion on fades and reverb tails. Dither adds a very small amount of random noise before the rounding, which turns that error into an even, steady hiss that is far less noticeable.
The rule is to dither once, and as the last step. Do it when you convert to 16-bit, not at every stage along the way. If the file stays at 24-bit or 32-bit float, no dither is needed.
Why not MP3
An MP3 throws away audio to get small. It does so in a way the ear usually tolerates, but it is permanent, and building on one means processing, pitching and compressing what is left. A producer who stretches or layers over an MP3 hears the loss get worse each time it is processed. Use MP3 for previews if you must. For anything someone else will build on, send the WAV, as the guide on sending loops puts it.
Settings to use
- Format: WAV, uncompressed PCM. Not MP3, not AAC.
- Sample rate: the rate your project runs at, almost always 44.1 or 48 kHz. Do not resample upwards before exporting.
- Bit depth: 24-bit. Use 32-bit float only if you are passing stems between DAWs and want nothing clipped on the way.
- Channels: stereo unless the whole loop is genuinely mono.
- Dither: off at 24-bit or 32-bit float. On, once, only if you must go down to 16-bit.
- Length: 10 to 60 seconds, which is what ColabLoop accepts. BPM by genre helps you pick a tempo that fits the style.
- Name it with the key and tempo, so whoever opens it knows what they have.
When it is ready, you can publish it on ColabLoop for other producers to build on. Nothing is sold: they ask, you decide, and credit is agreed before anyone starts.