avbParser

Documentation and parser for Avids proprietary .avb file format

Background

Avid's video editing software stores projects as .avb files. These are proprietary, undocumented binary files. Avid software can also export common MXF (Media Exchange Format) files. However, some project information like locators are not exported to MXF. For a project we needed to get this info out of the .avb files (in fact, we later gave up on the idea and went another way). Thus, Max Maurer and I tried to reverse-engineer the .avb file format.

Reverse engineering a .avb file

How do you start with reverse engineering a binary blob if you have never done this before? You open a hex editor, print out the first x bytes of the file and try to find patterns in it. Once you find them, you mark all occurences and try to make sense of them.
Some stuff we found out:

There are differences between files created on a Mac and on Windows. I assume that all Mac test files we had came from PowerPCs. Those are big-endian. PCs are little-endian. Thus, Mac files had frame identifiers (so we called them) like "RTTA", while PC files used "ATTR". Also, time stamps were reversed
You can easily recognize Unix timestamps. They are 4 bytes long. If they represent a reasonably recent time the first byte is 0x47 or 0x48
The frame identifiers were followed by additional bytes containing the length of the frame and its type.

We have only reverse-engineered the file structure. We do not know what each frame's content actually means. Sometimes it is quite obvious (strings, PCM wave data, etc). The more complex stuff (tracks, sequences, etc) is still unexplored, however.

avbParser - converting .avb to strange XML

Once we had some understanding of the file layout, Max modified a parser he had written for MP3 files to work with the .avb files. You can download the parser below. Source code is included in the Jar file. Parsing a file is done by calling

java -jar avbParser.jar file.avb

The code is quite clean. Thus, if you want to know the exact workings of the file format, please look into the code.

avbParser.jar as of 04.12.2008
sample .avb file (if you need larger files, just google them)
output of parsing the sample file

License & (No) Support

avbParser is licensed under a BSD license. We do not support avbParser and can not answer any questions regarding the file format (sorry, too much time has passed). If you find avbParser useful, please tell me what you are using it for.