MP3 and MPEG Audio MIME Type

What does audio/mpeg mean?

The audio/mpeg media type identifies an elementary stream of audio frames from MPEG-1 or MPEG-2. It covers all three MPEG audio layers, not only MP3. IANA lists .mp1, .mp2, and .mp3 as its extensions and deliberately lists no single magic number.¹

The familiar MP3 format is MPEG Audio Layer III. .mp2 normally means Layer II, while .mp1 normally means Layer I. Each layer uses a different compression design, so the number is a layer name rather than a sequence of compatible file revisions. RFC 3003 defines audio/mpeg broadly enough to carry any of them.²

Why is there no fixed MP3 magic number?

MPEG audio is organized as small frames, each with its own header. A stream may also begin with an ID3v2 tag rather than an audio frame. ID3v2 tags begin with the ASCII bytes ID3, or 49 44 33 in hexadecimal, but that identifies the metadata tag, not the MPEG audio itself.²³

Frame headers contain synchronization bits, version, layer, bit-rate, sample-rate, and channel information. Those values vary from file to file and sometimes from frame to frame. A robust detector therefore validates a sequence of frame headers and skips recognized tags instead of relying on a short fixed prefix. This is why the IANA registration says there is no magic number.¹

How MPEG audio and the MIME type developed

ISO/IEC published MPEG-1 Audio as ISO/IEC 11172-3 in August 1993. The standard defines coded high-quality audio and decoding at 32, 44.1, and 48 kHz for digital storage systems.⁴ MPEG-2 Audio followed in ISO/IEC 13818-3:1995 and extended the family.²

The media type came later. RFC 3003 registered audio/mpeg in November 2000 because MPEG-1 and MPEG-2 audio were already common on the internet but lacked one uniform MIME type. The RFC describes the payload as MPEG frames, possibly interspersed with non-MPEG data such as metadata.²

That history explains a common source of confusion: MP3 was already a well-established name before audio/mpeg became its standard web label. Servers should send an MP3 response as Content-Type: audio/mpeg, even though users almost always call the file an MP3.

How to open an audio/mpeg file

All current major browsers play MP3 through the HTML <audio> element. MDN records support in Chrome, Edge, Firefox, Opera, and Safari.⁵ VLC also supports MPEG Layers I, II, and III as well as ID3 tags.⁶

If a file does not play, check more than its extension:

• Confirm that the server sends audio/mpeg.
• Inspect the stream with a media probe to determine its actual MPEG version and layer.
• Check whether the download is complete and whether frame headers remain valid.
• Try a dedicated player such as VLC when an older .mp1 or .mp2 file is involved.

Support in Poly

Poly has a dedicated mapping from audio/mpeg and .mp3 to its MP3 audio type. The indexing pipeline decodes the file, records duration and common tags such as title, artist, album, genre, date, BPM, lyrics, and comments, and generates thumbnails when embedded cover art is available.

For successfully decoded MP3 files, Poly also:

• Creates time-based audio embeddings for similarity search.
• Transcribes speech into WebVTT and indexes transcript segments for full-text search.
• Opens the original file in the in-app audio player and seeks to a matching transcript moment.
• Makes the transcript available to file-reading tools and the agent.
• Allows the agent's multimodal backend to read MP3 audio directly, subject to an eight-hour direct-input limit.

There is one extension-level limitation. Poly's MIME table recognizes audio/mpeg, but filename-only inference maps .mp3 and does not map the registered .mp1 or .mp2 extensions. A Layer I or Layer II file supplied only by filename may therefore be classified incorrectly. Use an explicit audio/mpeg type, or convert a working copy to a format that the workflow recognizes.

Metadata, privacy, and safe handling

ID3v2 can store far more than a song title. Its frames can carry performer, copyright, lyrics, URLs, pictures, and private or application-specific information.³ Removing visible title and artist fields does not prove that every embedded frame or image has gone.

RFC 3003 also warns that private tags can carry arbitrary material and that executable content found there must not be run. An audio/mpeg object provides no built-in signing, encryption, or confidentiality.¹ Treat unexpected files as untrusted input, keep media parsers updated, and inspect metadata before publishing recordings.

Converting MP3 and other MPEG audio

Choose the destination based on how the copy will be used:

• FLAC or WAV when you need a lossless working file. Decoding MP3 into either format prevents another lossy step during editing, but cannot restore information already discarded.
• AAC in M4A or Opus for a compact delivery copy when your target software supports it.
• MP3 for broad playback compatibility when you are starting from a lossless master.

FFmpeg can identify the input from its contents and write a new file based on the output extension. Its documentation shows ffmpeg -i input.flac -id3v2_version 3 out.mp3 as an MP3 conversion example.⁷

Is audio/mpeg the same as audio/mp3?

Use audio/mpeg. It is the IANA-registered standards-tree type and is the value defined by RFC 3003.¹² Some software emits audio/mp3, but that spelling is not the registered type for ordinary MP3 files.

Also avoid confusing audio/mpeg with audio/MPA, which is registered for MPEG audio carried as an RTP payload. A saved .mp3 file and packetized real-time audio are different interchange contexts.