Convert OGG to FLAC

Editing is not one operation, it is a sequence: cut the top and tail, remove a cough, level the two speakers against each other, run a denoiser, export a rough for review, reopen it, fix the note the reviewer left, export again. Every one of those exports in a lossy format runs the audio through an encoder, and every pass discards a little more than the last.

The effect compounds and is easy to hear once you know what to listen for: sibilants get harsher, room tone develops a swirling quality, and reverb tails start to sound granular. A recording that has been round the loop five times as Vorbis is noticeably worse than the same recording exported once. FLAC in the middle of that chain removes the accumulation entirely — the intermediate saves cost nothing, and there is exactly one lossy encode, at the very end, into whatever the delivery format is.

It has to be said clearly because the opposite is widely assumed. If the Ogg holds Vorbis or Opus, the recording has already been through a lossy encoder and the FLAC preserves exactly what came out of it. The high frequencies that were rolled off are still rolled off; the artefacts that were introduced are now preserved perfectly.

What the conversion buys is a floor. From this file onwards the audio cannot get worse through handling, which is a different and genuinely useful property. Think of it as freezing the recording at its current condition rather than restoring it, and the decision about whether to convert becomes easy: yes if more passes are coming, no if this is the last stop.

Ogg is a container and the registry lists three codecs it carries: Vorbis, Opus and FLAC. That last one is not a curiosity. FLAC inside Ogg exists precisely so that a lossless stream can be carried in the Ogg framing, and it turns up in archival collections, in some Linux tooling and in a few download offerings.

If that is what you have, this conversion is a rewrap in substance: the same lossless audio moving from Ogg framing into a native FLAC file, with nothing lost at either end. Size tells you immediately which situation you are in — four minutes of FLAC-in-Ogg is twenty to thirty megabytes against three to five for Vorbis — and VLC's Codec Information panel confirms it in two clicks.

Both are lossless and the choice is not close. FLAC is roughly half the size for identical audio, which matters when a project holds twelve hours of interview material. It carries a checksum, so a file damaged by a failing disk or an interrupted copy announces itself instead of playing back wrong — the registry records that property and archivists rely on it.

It also keeps metadata that WAV cannot hold: Vorbis comments for the descriptive fields and PICTURE blocks for artwork, against RIFF INFO's seven fields limited to Latin-1 and no images. Audacity, Reaper, Ocenaudio, Adobe Audition, Hindenburg and DaVinci Resolve all read FLAC directly. Reach for WAV only where a specific piece of software or hardware refuses it.

FLAC compresses by predicting each sample from its neighbours and storing the error, which works well on a real waveform and less well on audio a psychoacoustic model has already reshaped. Decoded lossy material typically lands at 60 to 70 percent of its uncompressed size rather than the 50 to 55 percent a microphone recording achieves.

In practice: an hour of stereo at 44,100 Hz is about 600 MB uncompressed and roughly 400 MB as FLAC made from Vorbis. The same hour as the original Ogg was perhaps 55 MB. That is a large multiple, and it is the cost of the working copy — which is temporary. When the edit is finished and the delivery file exists, the FLAC intermediate has done its job and can be deleted.

This is the least lossy metadata path in the whole audio set, because both formats use the same tagging system. Ogg stores comments as uppercase key-value pairs — TITLE, ARTIST, ALBUM, ALBUMARTIST, TRACKNUMBER, GENRE, DATE, COMMENT, LYRICS — and FLAC stores exactly the same thing in its Vorbis comment block.

Cover art moves too: an image stored in Ogg as a METADATA_BLOCK_PICTURE comment becomes a proper PICTURE block in the FLAC. Custom keys a specific tool invented are the one category that can be dropped, since only the recognised fields are mapped. For an editing intermediate that is rarely a concern, but if a pipeline reads a custom field, extract it to a sidecar before converting.

Nothing is resampled and nothing is remixed. A 44,100 Hz stereo interview stays 44,100 Hz stereo; a 48,000 Hz field recording stays at 48,000; a mono lecture capture stays mono, at half the bytes of a pointless stereo duplicate.

The limit worth knowing is channel count: the registry records FLAC as carrying up to 8 channels, against Opus's 255. Ordinary stereo and 5.1 material is unaffected. A configuration no encoder will take falls back to two channels at 48,000 Hz rather than failing, so if you are working with an unusual multichannel capture, check the channel count of the first output before you build a session on it.

Decoding uses the codecs on the machine you are sitting at, which means the answer depends on your browser rather than on this page. Opus is universal because calls required it. Vorbis is widely but not uniformly supported, and FLAC inside Ogg is narrower again — the registry marks Ogg's support as patchy and that is an accurate description rather than a hedge.

Where the codec is absent, the conversion stops and reports that nothing in the file could be converted and that it may use a codec this browser cannot read, rather than writing an empty FLAC and calling it a success. If you meet that message, try the same file in a different browser before concluding the recording is damaged.

The recordings that arrive at this pair are often the kind that came with a consent form: an oral history, a research interview, a lecture recorded with the speaker's permission, a clinical or legal conversation. Those forms name who may hold a copy, and uploading a file to a conversion service creates a copy that nobody named.

Here the decode uses your own device's codec and the FLAC encoder is loaded into the tab and used nowhere else, so no request carries the audio. Files up to 100 MB are handled and a batch returns as a ZIP. As with every privacy claim on this site, the point is that you can check it rather than believe it: open the network tab of your developer tools, convert something, and look at what is sent.