User guide7 min read
KMID & KLyr XML Specification: Embedded MIDI Karaoke Lyrics
Comprehensive technical guide for KMID / KLyr format, Meta Event 0x01 extraction, Base64/zlib decompression, and XML schema (, , , ).
1. What is KMID / KLyr Format?
KMID (KAR / KLyr Embedded Format) is an advanced karaoke file specification that embeds song metadata, key signatures, and word-level XML lyric timing directly inside a standard MIDI (.mid) binary file. Unlike legacy 3-file NCN systems (.mid + .lyr + .cur), KMID packages everything into a single self-contained binary file, preventing missing lyric or sync errors during distribution.
2. Binary Extraction & Decompression Pipeline
Inside the MIDI track stream, KLyr data is stored within a Meta Text Event (Meta Type 0x01). The extraction and decoding pipeline follows four strict steps:
Extraction Steps
- 11. Header Matching: Search Meta Type 0x01 text events for the header identifier prefix
KLyrHdr1(orK1LyrHdr). - 22. Base64 Decoding: Extract the encoded string payload following the header and decode it into a binary
Uint8Arraybuffer. - 33. zlib/pako Inflation: Decompress the binary payload using zlib inflation algorithm (pako.inflate) to obtain raw XML bytes.
- 44. Charset Decoding (Windows-874): Decode XML bytes using Windows-874 / TIS-620 encoding to accurately preserve Thai vowels and tone marks.
3. Decompressed KLyr XML Schema Structure
The decompressed XML document contains two primary parent blocks: <INFO> for global song metadata and <LYRIC> for the line/word timing stream:
Decompressed KLyr XML Document Template
Decompressed KLyr XML Document Template
<?xml version="1.0" encoding="TIS-620"?>
<SONG>
<INFO>
<VERSION>1.1</VERSION>
<TITLE>ชื่อเพลง</TITLE>
<ARTIST>ชื่อศิลปิน</ARTIST>
<ARTIST_TYPE>M</ARTIST_TYPE> <!-- M: ชาย, F: หญิง, MF: ร้องคู่ -->
<KEY>Cm</KEY>
<TEMPO>120</TEMPO>
<ALBUM>ชื่ออัลบั้ม</ALBUM>
<AUTHOR>ผู้แต่งคำร้อง/ทำนอง</AUTHOR>
<GENRE>Pop</GENRE>
<RHYTHM>Disco</RHYTHM>
<CREATOR>NEXT_SYSTEM</CREATOR>
<COMPANY>ค่ายเพลง</COMPANY>
<LANGUAGE>THAI</LANGUAGE>
<YEAR>2026</YEAR>
<VOCAL_CHANNEL>9</VOCAL_CHANNEL>
</INFO>
<LYRIC>
<LINE>
<WORD>
<TIME>1920</TIME>
<TEXT>เนื้อ</TEXT>
<VOCAL>M</VOCAL>
</WORD>
<WORD>
<TIME>2160</TIME>
<TEXT>ร้อง</TEXT>
<VOCAL>M</VOCAL>
</WORD>
</LINE>
</LYRIC>
</SONG>4. Detailed Element Reference: <INFO> & <LYRIC>
Below is the technical specification of each XML element inside <INFO> and <LYRIC>:
<INFO> Node Element Reference
Identification
<TITLE> & <ARTIST>
Stores song title and lead artist name. Supports full Thai/English Unicode characters.
Music Theory
<KEY> & <TEMPO>
Defines default musical key (e.g. C, Cm, Dm, F#m) and tempo BPM. Allows player to display key and calculate transpositions.
Vocal Assignment
<ARTIST_TYPE> & <VOCAL_CHANNEL>
ARTISTTYPE defines M (Male), F (Female), or MF (Duet). VOCALCHANNEL specifies the guide vocal track (e.g., 9 for Drum/Guide channel).
<LYRIC>, <LINE>, and <WORD> Node Specification
- 1<LINE>: Groups words belonging to a single song verse line on screen.
- 2<WORD> -> <TIME>: Syllable start time stored in 24 PPQ reference ticks. Converted to real MIDI ticks using formula:
realTick = Math.round((time * ticksPerBeat) / 24). - 3<WORD> -> <TEXT>: Syllable character string (e.g. 'รัก', 'เธอ', ' Heart').
- 4<WORD> -> <VOCAL>: Optional vocal gender per word (M/F). Enables duet karaoke players to dynamically change font highlight colors (e.g. Blue for Male, Pink for Female) word-by-word!