Convert AAC to OGG

Ogg is a container and the registry lists three codecs it can hold — Vorbis, Opus and FLAC. Convention has attached `.ogg` to Vorbis for twenty-five years because that pairing came first, so a file with this extension is widely assumed to be Vorbis. This converter writes Opus, and a developer wiring up a fallback chain needs to know which one is arriving.

For the job on this page it is the better outcome. Opus was published as RFC 6716 in 2012 and outperforms Vorbis at every bitrate. The container is where the support question actually sits: Safari read nothing at all inside an Ogg wrapper until 18.4 in March 2025, so an `.ogg` source of either codec is a fallback for Chrome, Edge and Firefox rather than for everything. Among the clients that do open the container, Opus has the broader modern support of the two. What it will not satisfy is a build check or an ingest rule that names Vorbis specifically.

The gap is narrower than the folklore suggests and it has not closed. Chrome, Edge, Safari and released Firefox builds all decode AAC. What does not, reliably, is Chromium as packaged by several Linux distributions: those builds are compiled without the proprietary codec set for licensing reasons, and anything embedding an engine built the same way inherits the limitation.

The same reasoning produces open-source players and desktop applications that ship free codecs only. If your audience includes Linux desktops, self-hosted deployments or anyone running a distribution build, a second source costs one conversion and removes an entire class of silent failure. If your audience is phones and mainstream browsers, AAC alone is already sufficient and this page is solving a problem you do not have.

Worth stating rather than leaving to be discovered. The `ogg` and `opus` targets on this site both write an Opus stream in an Ogg container through the same encoder, so the audio produced by this page and by the AAC-to-OPUS page is identical. The extension is the entire difference.

Which one to pick is therefore a question about your tooling, not about the file. Asset pipelines, static-site plugins, upload forms and content-type checks frequently accept one name and not the other, and audio elements in the wild are more often written with `.ogg` because that convention is older. Pick whichever your build expects, and rename freely afterwards — the stream is legal under both.

A file with a `.aac` extension is an Audio Data Transport Stream: a run of AAC frames each carrying a short header, with no container around them. There is no index, no duration field and no metadata area. Files in this shape usually come out of a segmenter, a capture, an encoder configured for streaming, or a command-line tool asked for raw output.

That absence has one consequence for web work. Some players report the wrong length for an ADTS stream or refuse to seek in it, because the duration has to be estimated by scanning rather than read from a header. The Ogg produced here does carry that structure, which occasionally makes the fallback behave better than the primary — an argument for shipping the AAC in an M4A rather than raw, if you control the build.

A browser picking between source elements uses the declared type to decide what to attempt, and it does so before fetching anything. The registry records `audio/ogg` for this extension, and that is what belongs in the type attribute and in the header your server sends.

Getting it wrong produces the failure this exercise was meant to prevent: a client downloads a file it cannot decode, fails, and either falls through slowly or not at all. Serve the Ogg with a correct content type, list it after the AAC source rather than before it so capable clients take the better-supported file first, and test with a build that genuinely lacks AAC rather than assuming from a codec table.

The AAC already discarded parts of the recording permanently, and Opus now discards a further helping using a different model — spending part of its budget faithfully reproducing the first encoder's artefacts rather than the original sound. That is the standard cost of any lossy-to-lossy conversion and no setting removes it.

For a fallback that a minority of users will ever hear, Balanced at 64 kbps is a defensible engineering decision and keeps the extra asset small. For anything that will be a primary source on some platforms, High quality at 128 kbps is the right call: Opus at that rate is generous, and the point of giving the second encoder room is that it then contributes almost nothing of its own.

The three bands resolve inside the encoder: Small file writes 38 kbps, Balanced 64 and High quality 128. Anyone judging by MP3 numbers will read 64 kbps and expect something unpleasant. Opus at 64 kbps is a comfortable stereo music bitrate, and at 38 kbps speech is completely clear — the same codec runs messaging voice notes at 16 to 32.

For interface sounds, notification tones, short stingers and voice prompts, Small file is often correct rather than cheap, and on a page loading dozens of assets the difference is measurable in load time. For music beds and anything with sustained content, Balanced or High quality. The file sizes follow directly: a four-minute asset lands near 1.1 MB, 1.9 MB and 3.8 MB.

Ogg carries metadata as Vorbis-style comments — a plain list of uppercase keys with text values — and it is a perfectly good place to put a title, a licence or an attribution. The problem in this direction is upstream: a bare ADTS stream has no metadata area at all, so there is nothing to copy across and the result arrives with an empty comment header.

For build assets that is usually irrelevant, since the file name and the manifest carry the identity. Where it does matter — a downloadable clip, a Creative-Commons asset that must credit a source — the tags have to be written into the Ogg afterwards with a tagger, and doing it at this point in the chain is the right moment because they will then survive any further copying.

Opus is the only audio target on this site that fetches no encoder, because every browser had to implement it for calls, and your machine already decodes AAC. A conversion starts the moment a file lands and the network stays idle throughout, which you can confirm in developer tools while a batch runs.

Files are capped at 100 MB each and a hundred can be dropped at once, coming back as a ZIP — which is the usual shape of this job, since fallbacks are generated for a whole asset directory rather than one file. For unreleased product audio, the fact that nothing is transmitted is worth more than the convenience.