BFI: Difference between revisions

From MultimediaWiki
Jump to navigation Jump to search
m (→‎File Format: typical chunk signature)
(→‎File Format: undo: this is payload, not the chunk header!)
Line 44: Line 44:
The media frames started at file offset specified in bytes 4-7 of the header. Each frame is packaged inside a chunk with the fourcc 'IVAS' (possibly Interleaved Video and Sound). A IVAS chunk has the following payload:
The media frames started at file offset specified in bytes 4-7 of the header. Each frame is packaged inside a chunk with the fourcc 'IVAS' (possibly Interleaved Video and Sound). A IVAS chunk has the following payload:


   bytes 0-3      (  +0) unknown (video compression type? this is often 'IVAS')
   bytes 0-3      (  +0) unknown (video compression type?)
   bytes 4-7      (  +4) offset of the audio data (from chunk header!)
   bytes 4-7      (  +4) offset of the audio data (from chunk header!)
   bytes 8-11    (  +8) unknown (always zero)
   bytes 8-11    (  +8) unknown (always zero)

Revision as of 13:09, 27 March 2008

Multimedia-heavy games by Tsunami use a multimedia format with the extension .bfi. According to the README.TXT file on the game Flash Traffic: City of Angels, BFI stands for Brute Force and Ignorance. As described:

WHAT'S A BFI?

BFI stands for 'Brute Force & Ignorance'. This is the methodology we used to convert captured video to something which could be played back in a reasonably reliable manner on a wide range of currently installed machines. It simply means we don't think you should _have_ to buy a brand new 100Mz PCI bus Pentium machine with a true-color, high resolution, 'turbocharged' local-bus video card just to play a computer game. It's just too much jargon and frankly, it makes my head hurt a bit.

File Format

All multi-byte numbers are little endian. The general FourCC chunk format is as follows:

 bytes 0-3    chunk type
 bytes 4-7    chunk size, including 8-byte preamble
 bytes 8..    chunk payload

A BFI file begins with the following 960 (0x3C0) bytes header:

 bytes 0-3      (  +0) file signature: 'BF&I'
 bytes 4-7      (  +4) total file size
 bytes 8-11     (  +8) offset of the first frame in file
 bytes 12-15    (  +C) frame count
 bytes 16-19    ( +10) unknown
 bytes 20-23    ( +14) largest frame chunk size
 bytes 24-27    ( +18) unknown (perhaps video compression type)
 bytes 28-31    ( +1C) framerate
 bytes 32-43    ( +20) unknown
 bytes 44-47    ( +2C) video width
 bytes 48-51    ( +30) video height
 bytes 52-55    ( +34) first used palette entry index
 bytes 56-59    ( +38) number of used palette entries
 bytes 60-827   ( +3C) 256 RGB triplets (6-bit VGA palette values)
 bytes 828-831  (+33C) audio sample rate
 bytes 832-835  (+340) audio channels (?)
 bytes 836-959  (+344) unknown

The media frames started at file offset specified in bytes 4-7 of the header. Each frame is packaged inside a chunk with the fourcc 'IVAS' (possibly Interleaved Video and Sound). A IVAS chunk has the following payload:

 bytes 0-3      (  +0) unknown (video compression type?)
 bytes 4-7      (  +4) offset of the audio data (from chunk header!)
 bytes 8-11     (  +8) unknown (always zero)
 bytes 12-15    (  +C) offset of the video data (from chunk header!)
 bytes ..              unsigned, 8-bit PCM audio data (size = offset_of_the_video - offset_of_the_audio)
 bytes ..              encoded video data

Video compression

The video always encoded as palettized, 8-bit data. Single palette used for the entrie movie. Raster compression based on mixed LZ and RLE coding. Compressed stream begins with 4-byte value, representing final unpacked size of the data, then followed by compression codes.

Decompression algorithm:

read next byte of the input
code = top two bits of the byte
length = lower 6 bits of the byte

code is ...
 0 : normal chain
      if length is zero, read two more bytes of the length
      copy next (length) bytes of the input to the output
 1 : back chain
      if length is zero, read one more byte of the length AND two bytes of the offset
       else, read one byte of the offset
      copy (length) dwords from (output - offset) to the output
      (note, that data may overlap, so don't use memmove() function here, just bare by-dword copying!)
 2 : skip chain
      if length is zero, read two more bytes of the length
      if length still zero, finish the decompression
      leave (length) bytes of the output unchanged from the last frame
 3 : fill chain
      if length is zero, read two more bytes of the length
      color = next two bytes of the input
      fill (length) words of the output with (color)
  
repeat until stop code encountered or whole output generated

Games Using BFI