32-bit Float Recording on iPhone: What It Actually Does

Bit depth is just how each moment of sound gets written down as a number. 16-bit and 24-bit store that number on a fixed scale with a hard ceiling at the top, the point we call 0 dBFS. Go past the ceiling and the number cannot get any bigger. The peak is clipped flat and that part of the sound is gone for good.

32-bit float writes the number a different way. It works like a calculator holding very big and very small figures, with a value and a separate exponent that slides it up and down a huge range. The useful way to picture it is this. At any single instant you still get about the same detail as 24-bit, a clear window on the sound. The float part just lets that window slide across a scale so wide, around 1,500 dB in theory, that in normal use you never run out of room at the top or the bottom.

Because there is effectively no ceiling, a 32-bit float file will not clip in the digital world. A peak that shoots past 0 dBFS is simply stored with a bigger exponent instead of slamming into a wall. Pull the level back down afterwards and the peak returns exactly as it was, with no distortion.

That leads to the part that is truly worth having. You can raise or lower the level of a 32-bit float recording after the fact and it costs you nothing. A quiet take pushed up does not drag a layer of hiss up with it the way older formats would. A hot take pulled down stays clean. This is real, and it is the reason the format is worth using.

Here is the part the marketing skips. The number gets written as 32-bit float, but the sound still had to get into the phone through a microphone and an analog to digital converter first. Those set the real limits.

If the mic or the preamp in front of the converter distorts, that distortion is in the sound before it ever becomes a float number, and no bit depth can undo it. If the converter itself runs out of range at its input, same story. A 32-bit float file cannot rescue a sound that was already wrecked on the way in.

So on an ordinary single converter, which is what a phone has, 32-bit float does not give you a clip-proof input. It gives you a clip-proof project. Nothing clips once the sound is inside. Your summing and effects and gain moves all have endless headroom, and you can re-level anything later for free. That is the true and useful claim. The one to be suspicious of is anyone telling you the file format alone makes the recording itself impossible to ruin.

The recorders that really are clip-proof at the input, like the Zoom F3, do it in hardware. They run two converters at once, one set up for quiet detail and one for loud peaks, and stitch them together so nothing overloads no matter how loud it gets. The 32-bit float file is just the container big enough to hold that combined range. The two converters are the trick, not the file.

You can get that all the way into an iPhone, but only with the right gear and one thing people miss. The recorder has to actually output 32-bit float over USB, and the app has to record that float stream. A Zoom F3 does output float over USB and Reel will record it. A Zoom H1essential, by contrast, records 32-bit float to its own memory card but sends plain 16-bit over USB, so plugged into a phone you are not getting float at all. Always check the USB output spec, not just the box.

With the built-in mic you are on that single ordinary converter. iOS is happy to hand an app 32-bit float samples, and Reel records them, but it is a float container wrapped around the phone's own modest mic, not a clip-proof front end. You get every one of the in-the-box wins, no clipping in the mix, free re-leveling, clean overdubs, while the capture itself is still bounded by the little mic in the phone. That is plenty for catching ideas. It is just worth knowing where the limit really sits.

Does any of this beat plain 24-bit on sound? No. 24-bit already holds about 144 dB of range, more than any room or mic or converter will ever hand it and far more than your ears can hear. A 24-bit take with the gain set sensibly is every bit as clean as a 32-bit float one. There is no warmer, richer, better sound hiding in the bigger number.

What 32-bit float buys you is insurance against gain decisions, not better audio. When you cannot babysit the level, a sudden loud transient in a field recording, a band that jumps from a whisper to a wall of sound, a one-take moment you cannot redo, float means a peak you did not see coming does not cost you the take. When you are sitting down setting a level once in a quiet room, 24-bit is completely fine and the files are smaller.

Two practical notes. A 32-bit float file is about a third bigger than the same recording in 24-bit, so they add up faster. And while every serious DAW reads 32-bit float without blinking, some consumer and video apps still handle it badly, so if you are handing a file straight to a video editor a plain 24-bit or normalized version is the safer bet.

One last thing that trips everyone up at least once. Open a 32-bit float file in a DAW and any peaks that went over 0 dBFS can look clipped, flat-topped and alarming. They are not. The app just opened the file at full level. Pull the fader down and the peaks are all there, unharmed. That is the format working, not a damaged file.

Reel records 32-bit float up to 96kHz. With the phone's own mic that gives you a project you cannot clip and takes you can re-level and mix freely, which is the honest, useful half of the promise. Feed it a true 32-bit float recorder like the Zoom F3 over USB and you get the whole thing, genuinely clip-proof capture from mic to file. I would rather tell you exactly which half you are getting than sell you the word on the box.

Reel is the 32-bit float recorder I make and use, so if you want to see the workflow, here is a quick overview of how it works.