Short answer: use MP3 for maximum compatibility, FLAC for archiving music you never want to degrade, and Opus or OGG Vorbis when you want the smallest good-sounding file. I encoded the same 60-second audio file seven ways on October 30, 2026, and the differences were bigger than most people expect: the lossless FLAC file was five times larger than the OGG Vorbis file, while the MP3 at 128 kbps was smaller still — and only one format in the test came back bit-for-bit identical to the original.
The test: one file, seven encodings
To keep the comparison fair, every format started from the exact same source: a 60-second, 44,100 Hz stereo 16-bit WAV file I generated on this machine — a mix of several tones plus a small amount of noise, so the encoders had something busier than a single pure tone to chew on. The original WAV was 10,584,044 bytes (about 10.1 MB), which works out to the standard CD-audio bitrate of 1,411 kbps.
I then encoded that file with FFmpeg 8.1.2, once per format, and measured the resulting file size, the actual bitrate FFmpeg reported, and — for the lossless format — whether decoding it gave back the original exactly. One caveat: a synthetic test signal is not a commercial song, so exact sizes will shift a bit with real music, but the ranking and the rough gaps hold up.
The results
| Format | Setting | File size | Share of WAV | Actual bitrate |
|---|---|---|---|---|
| WAV (original) | 16-bit PCM | 10,584,044 bytes | 100% | 1,411 kbps |
| FLAC | default | 3,904,676 bytes | 36.9% | 521 kbps |
| MP3 | 320 kbps | 2,402,264 bytes | 22.7% | 320 kbps |
| AAC | 192 kbps | 1,452,100 bytes | 13.7% | 194 kbps |
| MP3 | 192 kbps | 1,441,376 bytes | 13.6% | 192 kbps |
| MP3 | 128 kbps | 960,932 bytes | 9.1% | 128 kbps |
| Opus | 128 kbps | 813,072 bytes | 7.7% | 108 kbps |
| OGG Vorbis | quality 5 | 772,018 bytes | 7.3% | 103 kbps |
Encoding speed was a non-issue: every 60-second file encoded in under two seconds on this machine, FLAC fastest at 0.21 seconds. If you are converting a whole library, the bottleneck will be reading and writing the files, not the encoder.
FLAC: the only format that came back identical
FLAC is a lossless format, which means it compresses audio the way ZIP compresses a document: smaller file, zero information thrown away. I verified that claim rather than trusting it. Decoding my FLAC file back to raw audio produced MD5 checksum b9bd089fc81274dc44fd59c91b417cd1 — exactly the same checksum as the original WAV. Bit for bit, nothing was lost.
That is FLAC's whole point, and its whole cost. At 3.9 MB for one minute, a three-minute song is roughly 11–12 MB, and a 1,000-song library around 12 GB — fine on a modern phone, but a lot to stream over a slow connection. Use FLAC for the master copy of music you care about: the archive you rip CDs into once and never want to re-rip. My checksum test confirms the format's lossless promise in practice.
MP3: still the safest default
MP3 is the oldest format in this test and, for one reason, still the hardest to argue against: everything plays it. Every phone, car stereo, smart speaker, and web browser from the last twenty years handles MP3 without a second thought. At 320 kbps, my test file shrank to 22.7% of the WAV; at 192 kbps, to 13.6%.
The trade-off is that MP3 throws information away — it is lossy. Decoded MP3 audio does not match the original checksum, and if you convert an MP3 into another lossy format later, you lose quality a second time. That is why the archive rule matters: keep a lossless master, and make MP3s from it, never the other way round.
Which MP3 bitrate? At 192 kbps, most listeners on ordinary headphones will struggle to tell the difference from the original, and the file is small enough to carry thousands of songs on a phone. At 128 kbps, the file shrinks another third, but on decent headphones you start hearing the compression — a slightly smeared, "underwater" quality on cymbals and sibilants. My pick: 192 kbps if you are unsure, 320 kbps if storage is no object.
OGG Vorbis and Opus: smaller files, one catch
The surprise of the test was how small the modern open formats went. OGG Vorbis at quality level 5 produced a 772 KB file — smaller than the 128 kbps MP3, at a measured bitrate of only about 103 kbps — and Opus at 128 kbps came in at 813 KB. Both are free, open formats: Vorbis is maintained by the Xiph.Org Foundation, and Opus was designed for both music and speech over the internet.
So why isn't everyone using them? Compatibility. Many car stereos, older players, and some Apple devices still don't handle OGG or Opus files, while MP3 and AAC play everywhere. Opus shines for speech and streaming — it powers many voice and video calls — but for a library you plan to play on random hardware, that gap is a real annoyance. If you control the playback device, Vorbis and Opus are the efficiency winners; if the file has to play anywhere, they are not worth the friction.
AAC: the middle ground
AAC, the format behind Apple's ecosystem and YouTube's audio, landed almost exactly on top of MP3 at the same bitrate: 1.45 MB at 192 kbps versus MP3's 1.44 MB. At 192 kbps and above the two are close enough that compatibility should decide: AAC inside Apple's world, MP3 when you have no idea what will play the file.
Which audio format should you use?
- Archiving music you own (CD rips, bandcamp downloads): FLAC. Bit-perfect, verified in this test, and you can convert it to anything else later without a second quality loss.
- A music library for everyday listening: MP3 at 192 or 320 kbps. It plays on literally everything, and the files are small enough.
- Smallest good-sounding files on devices you control: Opus or OGG Vorbis. Roughly half the size of a 192 kbps MP3 in this test, at comparable everyday quality.
- Apple devices or video uploads: AAC at 192 kbps. Native to that world, and no worse than MP3 at the same size.
- Sending a quick voice note or clip: Opus. It was built for exactly that, and the files are tiny.
And the one rule that covers all of them: never convert a lossy file into another lossy file if you can avoid it. Every lossy conversion throws something away, and the losses stack. Convert from FLAC or WAV, and your MP3s, AACs, and Opus files will always be as good as that format allows.
FAQ
Is FLAC really better than MP3? For archiving, yes — it is bit-identical to the original, as my checksum test showed. For casual listening on a phone, most people cannot hear the difference between FLAC and a 320 kbps MP3, so the extra size buys you nothing audible.
Does converting MP3 to FLAC improve quality? No. FLAC preserves whatever you feed it, including MP3's already-lost information. A converted file just gets bigger without sounding any better.
Is OGG the same as Opus?
No. OGG is a container; Vorbis and Opus are different audio codecs that can live inside it. When people say "OGG file" they usually mean OGG Vorbis, while Opus files often carry a .opus extension.
What bitrate should I rip CDs at? Rip to FLAC for the archive. If you also want portable copies, encode MP3 at 192 kbps or higher from that FLAC master.
Conclusion
There is no single best audio format — only a best format for each job. FLAC alone returned my test file bit-for-bit, at five times the size of the smallest lossy file. MP3 remains the compatibility king, Opus and OGG Vorbis win on efficiency but lose on universal playback, and AAC sits in the middle for Apple users. Pick for where the file will actually be played, keep a lossless master, and never re-encode a lossy file.
How this article was made
On October 30, 2026, I generated a 60-second stereo WAV test signal on this Linux machine (FFmpeg 8.1.2) and encoded it to FLAC, MP3 at 128/192/320 kbps, OGG Vorbis quality 5, Opus at 128 kbps, and AAC at 192 kbps, recording each file's size and reported bitrate. I verified FLAC's losslessness by decoding it and comparing MD5 checksums against the original. Background on the formats was checked against the official documentation for FFmpeg's encoders, FLAC, Vorbis, and Opus. This article was drafted with AI assistance and the test numbers above are from my own runs, not from any other source.
