What a sound file actually stores
Three numbers describe a recording, and only two of them apply to any one file. Sample rate is how often the sound was measured: 44.1 kHz for CDs, 48 kHz for anything that came out of video. Bit depth is how precisely each measurement is written down, and it means something only in an uncompressed format — 16-bit is the CD standard, 24-bit leaves headroom for further processing, 32-bit float is for mixing. Bitrate means something only in a compressed one: it is how many bits a second the encoder was allowed to spend.
That is why the sizes behave so differently. A WAV's size is arithmetic — rate times depth times channels — so a minute of 16-bit stereo at 44.1 kHz is about 10 MB whatever is in it, silence included. An MP3's size is whatever bitrate you chose: roughly 1.4 MB a minute at 192 kbps, and half that at 96. WAV is a working format rather than a listening one, and that distinction is the reason to convert in either direction.
Converting an MP3 to WAV restores nothing, and nothing ever will. The encoder threw detail away when the MP3 was made; a WAV built from it holds that damage at full resolution. What the conversion is for is stopping the loss going any further — from a WAV onwards, editing, trimming and re-saving cost nothing. The same applies to FLAC, which is lossless compression: a smaller way to store exactly the samples it was handed, not a way to improve them. An MP3 turned into a FLAC is a large file containing an MP3.
Every pass through a lossy encoder throws away more, and it never comes back. Converting AAC to MP3, OGG to MP3, or one MP3 into another is one such pass. Encoding at or above the source's own bitrate — which every page here shows you when you drop the file in — keeps that loss small; going far above it only spends bits storing the first encoder's mistakes. If you are going to do several things to a recording, convert to WAV first, do the work, and encode once at the end.
Cutting and joining need not re-encode at all. Compressed sound is stored in small frames — about 26 milliseconds for an MP3 — and Trim Audio copies those frames rather than decoding them, so a trimmed MP3 is bit for bit the MP3 it was, and a long recording cuts about as fast as it can be read. The price is that a copied cut begins on the nearest frame boundary, a few hundredths of a second either side of the point you picked; ticking the exact cut re-encodes to land on the sample you asked for. Merge Audio works the same way — files sharing a codec, sample rate and channel count are joined by copying, and files that differ are decoded and re-encoded to match the first.
Levelling is not loudness normalisation, and neither is compression. Adjust Volume measures the loudest moment in the file and lifts everything by that one amount, so the peak sits just under the ceiling. That is peak normalisation, and it is the right fix for a recording that is simply too quiet throughout. It is not the LUFS measurement that podcast and streaming platforms specify, which weighs perceived loudness across a whole programme — two files normalised to the same peak can still be obviously different in loudness. And lifting the whole file moves the quiet parts and the loud parts together, so it cannot even out a recording that swings between the two. That is dynamic-range compression: a different job, and not one these tools do.
Where a browser is genuinely the wrong tool
Running on your own device is what makes these private, and it buys you nothing at all in the way of processing. These tools convert, cut, join and set the level. They do not edit sound.
No repair and no shaping. There is no noise reduction, no hum or click removal, no EQ, no de-esser, no dynamic-range compression, no fades and no multi-track mixing. There are no crossfades between joined pieces either, which is why a cut in the middle of loud music can click. Audacity is free, runs on every desktop and does all of that properly — a hissy interview, or a recording with a fridge humming behind it, wants Audacity rather than this.
Peaks are measured; loudness is not. Adjust Volume reports the peak before and after and holds back anything that would pass the ceiling, but there is no true-peak limiter and no LUFS target. If you are delivering to a specification — −16 LUFS for a podcast, −14 for a streaming platform — you need a tool that measures loudness rather than peaks.
Tags do not travel. Titles, artists, album art and chapters are not carried across a conversion here yet: the sound comes through and the library metadata does not. For converting a music library, a desktop manager that rewrites the tags as it goes will save you hours of retagging.
One file, one click. Converting eight hundred voice memos is a shell loop around ffmpeg, which is free and will do in a minute what a person clicking would spend an afternoon on.
And some things no browser can do. Protected purchases — old .m4p files, Apple Music downloads, anything from a subscription — cannot be opened here, and stripping that protection is not something we help with. MP3 itself carries at most two channels at up to 48 kHz, so a surround or high-rate source is folded down on the way through.
Choosing between the tools that sound alike
Compress Audio vs WAV to MP3. They can produce the same file, and only one of them is the right way round. WAV to MP3 is a first encode from an uncompressed source, which is the cheapest encode there is. Compress Audio takes anything and re-encodes it to a bitrate you pick — from a WAV that is the same job, but from an MP3 it is a second lossy pass. It warns you when the bitrate you asked for cannot make the file smaller, which is what happens when you compress a 128 kbps MP3 at 192.
"Compress" means two things in audio, and only one of them is here. Compress Audio makes the file smaller by lowering the bitrate. Dynamic-range compression makes the quiet parts louder relative to the loud ones, which is what a podcast producer means by the word. Neither Compress Audio nor Adjust Volume does the second — Adjust Volume moves the whole file by a single amount.
AAC to MP3 vs M4A to MP3. Almost the same page. A .aac file is a bare AAC stream and a .m4a is AAC inside an MP4 box that can also hold artwork and chapters; both pages read both, and the sound inside is identical. The M4A page is written for iPhone voice memos, which are mono AAC at a low bitrate already and want 128 kbps mono on the way out.
MP3 to WAV vs FLAC to MP3. Opposite directions, and both one-way. MP3 to WAV makes a working copy you can edit without losing anything further; it does not improve the sound, and the file grows about tenfold. FLAC to MP3 makes a portable copy of a master you should keep, because an MP3 can never be turned back into the FLAC it came from.
Trim Audio vs Merge Audio. Trim cuts pieces out of one file — including pieces from the middle — and joins what is left. Merge puts separate files end to end in an order you set by dragging. An interview recorded in three parts with a cough in the second is both jobs: trim the cough out, then merge the three.