Audio Converter

How Browser Audio Conversion Works

Decoding is built in. Every browser ships decoders for the formats it can play, and the Web Audio API exposes them. Anything your browser can play, it can decode here.

Decoded audio is just numbers. Once decoded, the file is an array of floating point samples per channel. Trimming, mixing and fading are ordinary arithmetic on those numbers.

Resampling is done properly. Changing the sample rate renders the audio through an offline audio context, so the browser's own resampler is used rather than samples simply being dropped, which would introduce aliasing you can hear.

Nothing leaves the machine. There is no upload step, which removes the size limit, the queue and the question of what happens to your file afterwards.

How to Use This Converter

Drop the file and wait for the waveform. Decoding a long file takes a moment, and the waveform is drawn from the real samples once it finishes.

Trim against what you can see. The two sliders set the start and end, and the portion outside the selection is greyed out on the waveform so silence at either end is obvious.

Choose the output settings deliberately. Leaving the sample rate at the original avoids resampling entirely. Lowering it makes a much smaller file, which for speech costs nothing audible.

Listen before downloading. The result is playable in the page, so a fade that is too long or a normalisation that exposed background noise shows up before you save the file.

Why the Output Is WAV

Browsers decode but do not encode. The platform gives you decoders for every format it plays and encoders for almost none, so MP3 or AAC output would mean shipping a large encoder into your browser first.

WAV needs no encoder at all. The file is a 44 byte header followed by the raw samples, which is why it can be written directly and exactly.

It is lossless against what was decoded. No second round of lossy compression is applied, so nothing further is thrown away.

The cost is size. WAV is uncompressed, so expect roughly ten megabytes per minute of CD quality stereo. The estimate above tells you before you convert.

Audio Formats and What They Cost

MP3 is universal and lossy. Thirty years old, playable everywhere, and it discards audio it judges you will not miss. Fine for listening, imperfect as a master.

AAC and M4A are the better lossy option. More efficient than MP3 at the same bitrate, which is why they are the default in most modern devices and streaming.

FLAC is lossless compression. Roughly half the size of WAV with the samples recovered exactly, which makes it the sensible archive format when a player supports it.

Opus is the efficient modern one. Excellent at low bitrates and the default in most voice and conference applications, though support outside browsers is patchier.

Sample Rate and Bit Depth

Sample rate sets the highest frequency you can store. The ceiling is half the rate, which is why 44100 exists: it comfortably covers the top of human hearing at around 20000 cycles per second.

Speech needs far less. 16000 captures everything that matters in a voice and produces a file roughly a third the size. Telephone quality is 8000, and it sounds like it.

Raising the rate adds nothing. Resampling a 22050 recording to 48000 makes the file bigger and recovers no detail, because the information was never there.

Bit depth is dynamic range. 16 bit covers about 96 decibels, which is more than any listening environment. 32 bit float is useful as an intermediate for further editing rather than as a delivery format.

Normalising Against Loudness

Peak normalising scales to the loudest sample. The whole file is multiplied so its highest peak just reaches full scale. Nothing is compressed and the dynamics are unchanged.

It is not the same as loudness matching. One brief spike sets the ceiling for everything, so two files normalised the same way can still sound very different in volume.

It raises the noise floor too. Quiet recordings gain hiss along with signal, which is why listening to the result matters more here than anywhere else in the tool.

Fades solve a different problem. A cut at a non zero sample produces an audible click. A fade of even a tenth of a second removes it entirely.

When Mono Is the Right Choice

Speech is almost always mono. A single voice carries no stereo information, so two channels double the size for nothing.

Downmixing averages the channels. That is the correct approach for ordinary material, and it halves the file.

Watch for phase cancellation. If the two channels contain the same sound with opposite polarity, averaging them cancels it out. It is rare, and it is loud when it happens, so listen to the result.

Generation Loss and Quality

Converting lossy to lossless recovers nothing. An MP3 turned into WAV is a WAV of an MP3. It stops further loss, but what was discarded at encoding time is gone.

Each lossy generation costs something. MP3 to AAC to MP3 audibly degrades. Keeping a lossless master and encoding from it each time avoids the problem entirely.

Trimming and fading are lossless here. They operate on decoded samples and the result is written without compression, so no additional quality is lost in this tool.

Common Mistakes to Avoid

Upsampling in the hope of better quality. It only makes the file larger. The missing detail cannot be reconstructed.

Cutting without a fade. A hard cut mid waveform clicks. A very short fade at each end removes it.

Normalising quiet material without listening. The noise comes up with the signal, and a recording that was merely quiet can end up merely noisy.

Expecting WAV to be small. It is uncompressed by definition. If size matters more than fidelity, a lossy format is what you want, encoded from this WAV.

Frequently Asked Questions