DSF MIME Type

What does audio/x-dsf mean?

The audio/x-dsf media type identifies a DSD Stream File, a Sony container for Direct Stream Digital audio. A DSF file stores the one-bit pulse-density stream used by DSD rather than the multi-bit PCM samples found in WAV, AIFF, and FLAC files. Sony's specification defines .dsf as the filename extension and expands DSF as “dsd stream file.”¹

audio/x-dsf is a conventional, unregistered media type. It does not appear in the IANA audio media type registry.² The leading x- comes from an older practice for experimental or private names. Modern IETF guidance discourages creating new X- names, but established software mappings still use this value.³

How to identify a DSF file

A DSF file starts with a 28-byte DSD chunk. Its first four bytes are ASCII DSD , including the trailing space:

The following fmt chunk describes the version, channel layout, sample rate, sample count, and block size. A data chunk carries the DSD samples. An optional ID3v2 tag sits at the end of the file and can hold familiar library metadata such as title, artist, album, and artwork.¹

Sony's version 1.01 specification permits mono through 5.1-channel layouts and lists sample rates of 2.8224 MHz and 5.6448 MHz. Sample data is arranged in fixed 4,096-byte blocks per channel. Multi-byte numbers are little-endian, and one-bit samples are packed least-significant bit first.¹

Why is the sample rate so high?

DSD and PCM describe amplitude differently. PCM records a multi-bit value at each sample instant. DSD records a one-bit pulse-density stream at a much higher rate. More ones over a short interval represent a more positive signal, while more zeroes represent a more negative one. Sony describes DSD as a one-bit system fundamentally different from the sample-rate and bit-resolution model used by PCM.⁴

The large sample-rate number is therefore not directly comparable with a PCM sample rate. “2.8224 MHz, one bit” does not mean the same thing as “2.8224 MHz, 24-bit PCM.” DSD also relies on noise shaping, so conversion to PCM needs an appropriate low-pass filter and a chosen PCM sample rate and bit depth.

How to open an audio/x-dsf file

Use a DSD-aware application or playback device. Sony documents DSF playback on its HDD Audio Player at 2.8224 MHz and 5.6448 MHz. Sony's Hi-Res Audio Player also lists .dsf support at 2.8, 5.6, 11.2, and 22.4 MHz, subject to the connected output device.⁵⁶ FFmpeg can decode DSF files and is useful for inspecting or converting them.⁷

DSF is not a standard browser audio format. Current web-audio compatibility guidance covers formats such as AAC, FLAC, MP3, Opus, and Vorbis, but does not document DSF as a browser-playable container.⁸ A web application should provide a PCM or mainstream compressed alternative rather than expect an HTML <audio> element to decode the original .dsf file.

Playback can also depend on the output path. Some desktop players convert DSD to PCM in software. Native DSD output requires a compatible player, driver, transport mode, and DAC. If preserving native output matters, confirm the whole chain rather than relying on a “DSD supported” label for one component.

Support in Poly

Poly has an explicit audio/x-dsf type and maps .dsf filenames to it. That is enough to store, sync, organize, and find a DSF file by name and ordinary file properties.

Content processing is more limited. Poly's audio extraction and waveform paths use a decoder that does not include a DSF format reader. As a result, a DSF file does not currently yield duration and tags, an audio-similarity waveform, or the decoded snippets required for automatic transcription. Without a transcript, full-text search cannot find spoken words inside the recording.

The interface still routes recognized DSF files to the audio player. That player uses the original media URL, so playback depends on the browser decoding DSF, which mainstream browsers do not normally do. Poly does not generate a browser-friendly playback copy for this format.

The Poly agent has two possible routes for supported audio: a stored transcript or direct audio input. DSF currently has neither. The backend's direct-audio allowlist covers WAV, MP3, AIFF, AAC, Ogg, and FLAC, but not DSF.

Converting DSF to FLAC or WAV

FFmpeg can decode DSF and convert its DSD stream to PCM. A simple FLAC conversion is:

ffmpeg -i input.dsf output.flac

Choose explicit sample-rate, bit-depth, and filtering settings when a production workflow requires them. DSD-to-PCM conversion changes the sample representation and is not bit-preserving. Keep the original .dsf if you may need native DSD playback or an archival copy.

Use the destination that suits the next step:

• FLAC for lossless PCM storage with tags and broad software support.
• WAV for uncompressed PCM interchange and editing.
• AAC, MP3, or Opus for smaller delivery files where lossy compression is acceptable.

Converting DSF to another DSD container can preserve DSD samples when the tool performs a lossless remux, but metadata and bit-packing conventions still need attention. Merely renaming the extension never converts the file.

Is DSF the same as DSDIFF?

No. Both can store Direct Stream Digital audio, but they are different containers.

The Philips DSDIFF specification was designed for storage and interchange of DSD and losslessly compressed DST material.⁹ Sony's later DSF layout is convenient for individual consumer tracks because it includes a direct pointer to an optional ID3v2 tag. Support for one container does not guarantee support for the other, even when both carry DSD audio.