CD-text data is defined in a scattered manner between MMC-3 and Sony documentation. The below uses GNU libcdio's description.[4]
On the lowest level, CD-text is stored in 18-byte "pack" units; this part is defined in MMC-3 Annex J. Each pack consists of 4 bytes of header (type indicator, track number reference, sequential counter, block number and character position indicator [BNCPI]), 12 bytes of payload, and 2 bytes of CRC. The type indicator ranges from 0x80 to 0x8F, the 13 defined values being:[5]
More information Type, Keyword ...
CD-Text keywords
Type | Keyword | Description | Section | Format |
0x84 | ARRANGER | Name(s) of the arranger(s) | Any | Character |
0x83 | COMPOSER | Name(s) of the composer(s) | Any | Character |
0x86 | DISK_ID | Disc Identification information | Disk | Binary |
0x87 | GENRE | Genre Identification and Genre information | Disk | Binary |
0x8e | ISRC | International Standard Recording Code of each track | Track | Character |
0x85 | MESSAGE | Message from the content provider and/or artist | Any | Character |
0x81 | PERFORMER | Name(s) of the performer(s) | Any | Character |
0x82 | SONGWRITER | Name(s) of the songwriter(s) | Any | Character |
0x80 | TITLE | Title of album name or track titles | Any | Character |
0x88 | TOC_INFO | Table-of-content information | Disk | Binary |
0x89 | TOC_INFO2 | Second table-of-content information | Disk | Binary |
0x8e | UPC_EAN | UPC/EAN code of the album | Disc | Character |
0x8f | SIZE_INFO | Size information of the block | Any | Binary |
Close
The BNPCI is used to define information that does not fit in one pack. This can be text or binary data. The BNCPI also indicates whether the text is single-byte or double-byte data in the top bit. This determines how null-terminated strings are defined โ one or two bytes of 0x00.[4] (Note: the DBCS mode is rarely, if ever, used. Its special null handling is not necessary for computer DBCS code pages, as they are "hybrid" with ASCII and compatible in the NUL behavior. UTF-16 could be the intended use.)
For block types listed above as "character" (per MMC-3), the payload is a simple null-terminated string. (MMC-3 is written confusingly here โ it describes the encoding as "ASCII" in the pack type table despite mentioning the BNCPI flag modifying its behavior later.) The descriptions of the binary fields are vague, but the developers of GNU libcdio has either matched them to sections of MMC-3 or written new descriptions based on Sony's sample.[4]
Another layer of encoding specification is found at this payload level, in the SIZE_INFO block. Here the first byte may be used to indicate the encoding, ASCII, Latin-1, or "MS-JIS". This is supported by the original Sony authoring tools.[4]