- Website: http://www.thbeck.de/Tak/Tak.html (German)
TAK (Tom's verlustfreier Audio Kompressor), formerly known as YALAC (Yet Another Lossless Audio Codec) is a lossless audio codec developed by Thomas Becker. As of January 2006, the codec is at the 1.0 release stage. Source code is not available.
- Alpha testing thread from hydrogenaudio.org
- Beta testing thread from hydrogenaudio.org
- 1.0 release thread from hydrogenaudio.org
- Binary download link
Integers are little-endian. 24-bit CRC is employed for checking data integrity.
TAK file consists of signature ("tBaK"), obligatory metadata and compressed audio data. APE tags may be present at the end of file.
Metadata consists of series of objects, each object has 32-bit header.
7 bits - object type 1 bit - reserved 24 bits - object size
Metadata object types:
0x00 - end of metadata 0x01 - stream info 0x02 - seektable (obsoleted in TAK 1.1.1) 0x03 - original file data 0x04 - encoder info 0x05 - padding (since 1.0.3) 0x06 - MD5 checksum (since 1.1.1) 0x07 - last frame information (since 1.1.1)
Non-empty metadata object ends with 3 bytes of object data CRC.
Stream information consist of:
Encoder info Frame size information Audio format
6 bits codec type 4 bits encoder profile (= encoder preset at least for versions <= 1.1.2)
Known codec types:
0 Integer 24 bit (TAK 1.0) 1 Experimental! 2 Integer 24 bit (TAK 2.0) 3 LossyWav (TAK 2.1 Beta) 4 Integer 24 bit MC (TAK 2.2)
frame size information (32 bits, see below)
Frame size information
4 bits frame size type 35 bits number of samples in whole stream
Possible frame types:
0 94 ms of audio 1 125 ms 2 188 ms 3 250 ms 4 4096 samples 5 8192 samples 6 16384 samples 7 512 samples 8 1024 samples 9 2048 samples 10 obsoleted in 1.1.0, invalid, never used 11 obsoleted in 1.1.0, invalid, never used 12-14 reserved 15 frame contains number of samples stored in next 35 bits (not supported yet)
3 bits data type (should be 0 = PCM) 18 bits sample rate - 6000 5 bits sample bits - 8 4 bits audio channels - 1 (5 bits in 1.1.1 version, last bit is never used anyway because older encoders did not supported multi-channel input)
1 bit is extension present
(>= 2.2.0 only)
if extension is present: 5 bits valid bits per sample 1 bit is speaker assignment present if speaker assignment present:
6xCh bits speaker assignment for each channel
0 none 1 FRONT_LEFT 2 FRONT_RIGHT 3 FRONT_CENTER 4 LOW_FREQUENCY 5 BACK_LEFT 6 BACK_RIGHT 7 FRONT_LEFT_OF_CENTER 8 FRONT_RIGHT_OF_CENTER 9 BACK_CENTER 10 SIDE_LEFT 11 SIDE_RIGHT 12 TOP_CENTER 13 TOP_FRONT_LEFT 14 FRONT_CENTER 15 FRONT_RIGHT 16 BACK_LEFT 17 BACK_CENTER 18 BACK_RIGHT
Original file data
Called "simple wave data", this contains trailer and footer from original file to allow perfect reconstruction:
0-2 trailer size 3-6 footer size trailer data footer data
24 bits encoder version (8 bits per each version component) 4 bits preset 2 bits evaluation 2 bits reserved
Last frame information
40 bits frame position relative to the beginning of audio data 24 bits frame size
TODO: header order, padding, CRC positions
16 bits syncword (0xA0FF little-endian) 1 bit last frame 1 bit frame contains stream information + extra info (see "stream information" above and "extra info" bellow) 1 bit frame contains metadata (unsupported) 21 bits frame number 14 bits sample count - 1 (only for the last frame) 2 bits padding (only for the last frame) if stream information is present: stream information extra info
1 bit previous frame position is present (currently always set to 0) 5 bits reserved (should be 0) if previous frame position is present: 25 bits offset to the previous frame header from the end of this header