Convert MP4 to FLAC

FLAC has been a lossless codec since 2001 and was standardised as RFC 9639. It reconstructs every sample it was given, bit for bit, and it carries a checksum so a damaged file tells you it is damaged instead of playing quietly wrong. Those are real properties and they are why it is the right format for a master.

None of it helps with what is inside an MP4. The soundtrack there is AAC, which is lossy: parts of the signal were discarded when the video was made and they are not in the file to be recovered. Encoding what remains as FLAC preserves it perfectly and adds nothing. The word people hear is "lossless" and the word that applies is "unchanged".

The audio in an MP4 typically occupies one to two megabytes a minute. The same audio as FLAC lands somewhere around five to ten megabytes a minute, because it is now describing every sample rather than a compressed approximation of them. A one-hour recording goes from roughly ninety megabytes to somewhere between three hundred and six hundred.

FLAC is doing its job in that figure — it is roughly half the size of the same audio as an uncompressed WAV. The point is only that the increase buys nothing audible. You are paying for the ability to keep the file unchanged through future work, which is a genuine thing to pay for and a different thing from quality.

When the audio is going to be handled repeatedly. Restoration work — de-noising a bad recording, removing hum, levelling an interview recorded in a room with a fan — means opening, altering and saving the file many times. Every one of those cycles on a lossy file decodes and re-encodes, and after half a dozen the damage is audible.

A FLAC intermediate breaks that chain. Extract once, work in FLAC for as long as the work takes, and encode to a lossy delivery format once at the end from the finished result. The same argument covers transcription pipelines, forced alignment and any analysis that measures the waveform rather than listening to it.

Filing a lecture, a podcast episode or a piece of music from a video into a listening library. The FLAC will be five to ten times the size of an M4A of the same recording and will sound the same on any equipment, because the information that separates them was already removed before you started.

The instinct behind it is sound — keep the best copy — and the way to act on it here is to keep the MP4. That is the best copy. A FLAC made from it is a larger file containing strictly less than the original, since the original also has the picture.

Both are lossless and both are exact, so the choice is about everything around the samples. FLAC is roughly half the size, carries a checksum that detects corruption, and has proper fields for title, artist and album. WAV has none of those and is a plain box of samples.

WAV wins only on acceptance. Some older audio software, some transcription tools and some broadcast systems read WAV and nothing else, and there the decision is made for you. Where both are accepted, FLAC is the better file and the difference in size on a long recording is measured in hundreds of megabytes.

A FLAC file holds one audio stream. A video carrying an original language, a dub, a commentary or separate microphone feeds from a multi-camera shoot produces a FLAC containing the first of them, with the others discarded and no way to choose.

For material from a phone, a screen recorder or an ordinary export that is the only track and the question never arises. For anything professionally delivered, look at what is in the file before building on the result — the wrong feed is easy to notice late and expensive to fix then.

The limit applies to the file you drop in, and the FLAC that comes out will be much larger than the audio inside it was. A 100 MB video of a long talk can produce a FLAC well past that figure, built in the browser’s memory before it reaches your disk.

That is the honest ceiling for browser-based lossless extraction, and past it the right tool is a desktop one that can stream to disk rather than hold a file in a tab. Saying so is better than starting a conversion that runs out of memory two-thirds of the way through a three-hour recording.

The FLAC encoder is loaded on demand the first time you ask for one, rather than being downloaded by everybody who visits the site. Once it is there, the work is decoding the AAC and running lossless compression over the samples, which is arithmetic rather than perceptual analysis and is quick.

There is no quality setting, and that is not an omission. A lossless encoder has nothing to trade — every setting would produce the same audio and differ only in how hard it worked to make the file slightly smaller. The absence of the control is the format being honest about itself.

The decode and the encode both run inside the browser tab on your own processor, so the video is never uploaded. There is no queue, no account and no daily count, and you can check the claim by watching the network tab while a file converts.

The material this page attracts makes that worth stating plainly. People extracting lossless audio are usually working on something they care about or something they are responsible for — an oral history, a legal recording, a piece of music not yet released — and the most reliable way to keep a recording private is not to send it anywhere.