RL2

• Extension: RL2
• Samples: http://samples.mplayerhq.hu/game-formats/voyeur-rl2/

The PC version of the FMV-based adult game "Voyeur", developed by Entertainment Software Partners, heavily uses this custom format for its videos. This article only describes the format version found on the PC version of the game; no CD-i version details is known so far. In general, video is compressed using a simple RLE scheme and audio is carried as uncompressed, unsigned, 8-bit PCM data.

Trivia: According to Wikipedia, this game caused ESRB rating system to be invented.

File Format

+  0  dword  Header                -- "FORM"
+  4  dword  BackSize              -- size of the background frame
+  8  dword  Signature             -- "RLV2" or "RLV3"
+  C  dword  DataSize              -- size of the data past this point BIG-ENDIAN
+ 10  dword  NumFrames             -- number of frames
+ 14  word   Method                -- encoding method. ignored, zero
+ 16  word   SoundRate             -- sound sample rate in some obscure format. zero means no sound
+ 18  word   Rate                  -- sound sample rate in Hz
+ 1A  word   Channels              -- number of sound channels
+ 1C  word   DefSoundSize          -- size of the single sound chunk. see notes below
+ 1E  word   VideoBase             -- initial drawing offset within 320x200 viewport
+ 20  dword  ClrCount              -- number of used colors
+ 24  RGB    Palette[256]          -- 256 RGB triplets. full 8-bit entries
-- if Signature == "RLV3" AND BackSize <> 0
  +324 byte  BackFrame[BackSize]   -- encoded background frame. ignore VideoBase during it's decoding
--
+xxx  dword  ChunkSize[NumFrames]  -- complete size of the chunk for each frame (audio+video)
+yyy  dword  ChunkOffs[NumFrames]  -- offset of the each frame chunk from the start of file
+zzz  dword  SoundSize[NumFrames]  -- size of the audio portion in the frame chunk

-- for each frame --
+xxx  byte   Audio[SoundSize[frame_nr] & 0xFFFF]  -- raw 8-bit audio
+yyy  byte   Video[...]                           -- compressed video stream

• All numbers are little-endian (except for the DataSize).
• If sound is present (SoundRate <> 0) then fps = Rate / DefSoundSize. Otherwise, fps = 11025 / 1103.

Video Decoding

The video is always encoded in 320x200 resolution. If no background frame is present, video decoding operates as follows:

repeat
  tag = next byte of input
  tag < 0x80 : paint pixel with (tag)
  tag = 0x80 : length = next byte of input
               if length is zero - finish
               fill next (length) bytes of output with (zero)
  tag > 0x80 : length = next byte of input
               fill next (length) bytes of output with (tag & 0x7F)

For background-frame based decoding, the algorithm is slightly altered (you need to decode BackFrame using the above scheme first):

copy pixels from background-frame until videobase is reached
repeat
  tag = next byte of input
  tag = 0x00 : copy pixel from BackFrame
  tag < 0x80 : paint pixel with (tag | 0x80)
  tag = 0x80 : length = next byte of input
               if length is zero - finish
               copy (length) pixels from BackFrame
  tag > 0x80 : length = next byte of input
               fill next (length) bytes of output with (tag | 0x80)
copy pixels from background-frame until the output frame is complete