AUD · Audio tools

BWF Metadata Editor (bext, iXML, LIST-INFO)

Shareable link

Settings are written to the URL as you change them. Nothing differs from the defaults yet.

Oxyd To cut, arrange and export the whole episode with chapters and a loudness target: Edit the episode in Oxyd

Edit the metadata a Broadcast WAV carries beside its audio: who made the file, when, where it sits on a timeline, and how it was coded. The audio is copied into the saved file byte for byte.

The bext chunk

A Broadcast WAV is a WAV file with a bext chunk, defined in EBU Tech 3285. The chunk has a fixed layout, so every field has a hard length:

Field Size Holds
Description 256 characters Free text; archives put the file’s identifiers here
Originator 32 characters The body responsible for the file
Originator reference 32 characters An identifier issued by the originator
Origination date 10 characters yyyy-mm-dd
Origination time 8 characters hh:mm:ss
Time reference 64-bit integer Samples since midnight
Coding history Unlimited One line per generation of the audio

The counters beside Description, Originator and Originator reference show how much of each field is used. Text past the limit is cut on save, and the check list says so first.

Open a plain WAV and the fields are empty. Fill any of them in and the saved file gains a bext chunk in front of the audio.

Time reference

The time reference is what makes a file land at the right place when it is imported into a session. A recorder running on time-of-day timecode that starts a take at 10:00:00 writes 1728000000 into a 48 kHz file, which is 36,000 seconds of samples.

Type either form. 01:00:00.000 and 172800000 are the same value in a 48 kHz file, and the line under the field shows both. Changing the sample rate of a file changes the meaning of this number, so check it after a sample rate conversion.

Coding history

Each line records one stage in the life of the audio, as comma-separated pairs:

A=PCM,F=48000,W=24,M=stereo,T=Studer A820; transfer to file

A is the coding algorithm, F the sample rate, B the bit rate for lossy codecs, W the word length, M the mode (mono, stereo, dual-mono, joint-stereo) and T free text. Lines are saved with the carriage return and line feed the specification asks for, whatever your keyboard typed.

Loudness fields

Version 2 of the chunk added five loudness values: integrated loudness, loudness range, maximum true peak, maximum momentary and maximum short-term loudness. Each is stored as a whole number of hundredths, so -23.04 is kept exactly and a third decimal is rounded away.

Leave a field empty and it is written as “not set”, which is a different thing from zero. Filling one in raises the chunk to version 2 on save.

LIST-INFO and iXML

LIST-INFO is the older RIFF tag list: title (INAM), artist (IART), creation date (ICRD), comment, copyright, archival location and others. Media players read these and ignore bext. Entries this form does not show are kept in the saved file.

iXML is an XML document written by field recorders, with the project, scene, take, tape, track names and timecode rate. It is shown as text and saved exactly as edited. Malformed XML blocks the save: a reader that parses iXML has nothing to fall back on when the document does not parse.

Checks

The check list reads the file’s structure and every field as you type:

  • ERR: a truncated chunk, a missing format or data chunk, non-ASCII text in bext, a field over its length, an audio checksum that does not match.
  • WARN: a header size that disagrees with the file, a date or time in the wrong form, a coding history line that is not A=, F=, B=, W=, M=, T= pairs, a partial last sample frame.
  • NOTE: what the FADGI guideline strongly recommends and the file lacks: Originator, Originator reference, Description and Origination date, and an archival location (IARL) matching Originator whenever LIST-INFO is used.

A header size that was never filled in is the mark of a recording that stopped without closing its file. Saving writes the true size, which is often enough for a DAW to open it again.

Audio MD5

The MD5 chunk stores a checksum of the data chunk alone. Because it leaves the metadata out, it stays valid through any number of metadata edits, and it changes if a single sample does.

Verify computes the checksum of the audio and compares it with the stored one. Tick the embed box to write the checksum into the saved file. Hashing reads the whole of the audio, so a multi-gigabyte file takes as long as the disk needs to read it.

What saving writes

The saved file keeps the source’s chunks in their original order. Edited chunks are replaced where they stood, new ones go in front of the audio, and every chunk the editor does not handle (cue , smpl, axml, cart, _PMX and the rest) is copied unchanged.

An RF64 or BW64 file stays in that form with its ds64 sizes rewritten. A RIFF file stays RIFF.

Limits

One file at a time, with no CSV import. axml, cart and XMP chunks are preserved and not editable here. An RF64 file whose ds64 table lists chunks other than the audio as larger than 4 GB can be read and not saved.

Frequently Asked Questions

It opens a WAV file, shows its chunk layout and the metadata in its bext, LIST-INFO, iXML and MD5 chunks, lists what breaks EBU Tech 3285 or the FADGI embedding guideline, and saves a copy with your edits. A plain WAV with no bext chunk gets one when you fill in a field.

RIFF WAV, Broadcast WAV (.wav or .bwf), RF64 and BW64, at any bit depth, channel count and size. Sony Wave64 (.w64) is a different container and is not read.

No. The saved file is the edited metadata chunks plus the original audio bytes copied across. The MD5 of the data chunk is the same before and after, and the editor can compute it on both files to show that.

Yes. Only the chunk headers and the metadata chunks are read into memory, so an RF64 file of any size opens in the time a small one does, and saving streams the audio from the original file into the copy.

The number of samples between midnight and the first sample of the file, stored as a 64-bit count. Divided by the sample rate it gives the time of day the recording started, which is how an editor places a clip back at its original timecode. Type it as a sample count or as hh:mm:ss.mmm.

EBU 3285 defines the bext text fields as ASCII. Readers disagree about anything else, so a non-ASCII character is written as a question mark. LIST-INFO and iXML are written as UTF-8 and keep accents.

No, one file at a time. Copy report as JSON gives the layout, every field and the check results for the file that is open.

Explore Our Tools

Browse all tools