Convert WAV to FLAC

This is the rare conversion with no trade in it. FLAC predicts each sample from the ones immediately before it, stores only the difference, and packs that difference efficiently. Decoding runs the prediction backwards and reconstructs the original stream exactly — not something that sounds the same, the same numbers. A WAV of 10 MB a minute typically becomes 5 to 6 MB a minute.

That is a much weaker result than MP3 or Opus manage, and it is weak for a reason worth understanding. A lossy encoder gets to decide that a quiet cymbal behind a loud snare is inaudible and delete it. FLAC is not allowed to decide anything; it can only find redundancy that is genuinely there. Fifty percent is what honest compression of audio looks like, and everything beyond it is somebody throwing away part of the recording.

The compression ratio is a property of the recording, not of any setting. Sparse, quiet, predictable material compresses hard: an unaccompanied voice, a solo piano, a field recording of a room tone, an interview with one microphone. Forty to forty-five percent of the original is normal for that material, and speech recorded in mono can go further still.

Dense, loud, heavily limited material compresses badly. A modern mastered rock or electronic track has been pushed close to full scale for most of its length, which leaves the predictor very little to work with, and 65 to 70 percent is a realistic result. Noise is the worst case by definition — genuine white noise is incompressible, and a hissy cassette transfer behaves more like noise than like music.

A FLAC stream header holds an MD5 of the decoded audio, computed at encode time. That single field is the difference between a copy and an archive: years later a decoder can reconstruct the samples, hash them, and compare, so a drive that silently corrupted a block announces itself instead of quietly playing a click. WAV has no equivalent — a bit-rotted WAV is still a structurally valid WAV.

This is the argument for FLAC that has nothing to do with disk space, and it is the one that matters for material you cannot re-record. Digitised tapes, live sets, interviews, anything with one source and no second copy. The size saving pays for the migration; the checksum is what you actually bought.

FLAC specifies a maximum of eight channels. That covers mono, stereo, quadraphonic and 7.1, which is every layout most people will ever hold. WAV goes further — its extensible header names eighteen speaker positions — so a genuinely wide multichannel WAV has nowhere to go here. In practice this bites on ambisonic and immersive material rather than on music, and a stereo mix, a stem or a CD rip never comes near it.

Bit depth is decided by the source rather than by a setting: the encoder writes 16- or 24-bit output according to the sample format it is handed. That is the behaviour you want from an archive format, and it is worth contrasting with the WAV and AIFF targets on this site, which are written as 16-bit PCM regardless of what went in.

FLAC has a real metadata system: named text fields for title, artist, album and anything else you invent, plus a dedicated block for cover art. WAV has an optional INFO chunk that a great many recorders, editors and export dialogues never write at all. Converting cannot invent what the source did not carry, so an untagged WAV becomes an untagged FLAC.

The useful consequence is that this is the point in the chain where tagging finally sticks. Tags written into a FLAC travel with it into every player and every subsequent lossless conversion, which was never reliably true of a WAV. If the files are going into a library, tag them once now rather than after the next migration.

The conversion preserves what it is given, which is a guarantee about the process and not about the audio. A WAV that was itself decoded from an MP3 contains exactly what that MP3 contained, and wrapping it in FLAC preserves the damage perfectly. The file will be lossless and the recording will not be.

This matters because WAV is the format everything exports to, so a WAV in a folder is not proof of provenance. If a file came off a CD, out of a recorder, or straight from a DAW bounce, it is a master and FLAC is the right home for it. If it came from a download of unknown origin, converting it to FLAC produces a large file with an authoritative-sounding extension and no more information in it than before.

CD-quality stereo is 44,100 samples a second at two bytes each across two channels — 176,400 bytes every second, or about 10 MB a minute, whatever the music is doing. The free ceiling here is 100 MB per file, which works out at a little under ten minutes of stereo before a file is turned away.

That is enough for a song, a stem, a voice memo or a scene, and not enough for a concert or a two-hour interview. For a long recording, split it in an editor first, or convert the source with a desktop encoder; saying so is more useful than letting a 900 MB drop fail after it has been read.

Some browsers expose a FLAC encoder through WebCodecs and some do not. Where one exists it is used directly; where it does not, a WebAssembly FLAC encoder is downloaded the first time you convert to FLAC and never on any other page. That is why a visitor converting a spreadsheet on this site downloads none of it.

Either way the encoding happens inside the tab, on your processor, which is the only reason this pair can be offered without a queue or an account. You can open the network tab during a conversion and watch: the encoder comes down once, and the audio goes nowhere.

Drop up to a hundred WAVs at once and each is encoded on its own, then handed back together as a ZIP. This is the ordinary case for this pair rather than an edge case — recorders, DAWs and rippers produce takes, stems and tracks in sets, and nobody archives one file at a time.

The work is real processor time on your own machine rather than bandwidth, so a large batch is worth starting and leaving. Since nothing is uploaded, the only limits are the hundred files and the 100 MB each, and neither of them is a queue behind somebody else.