Dump subtitles

• Dump title 1 - 6

  #! /bin/sh -e
  for title in `seq 1 6`; do
  	mplayer dvd://${title} -dumpstream -dumpfile title${title}.vob
  done 

• Extract raw subs from vob with transcode

  #! /bin/sh -e
  for title in `seq 1 6`; do
  	tccat -i title${title}.vob -L| tcextract -x ps1 -t vob -a 0x21 > title${title}.sub
  done

• Convert to vobsub (needs VTS_01_0.IFO from DVD)

  #! /bin/sh -e
  for title in `seq 1 6`; do
  	subtitle2vobsub -o itcrowd${title} -i VTS_01_0.IFO -a 2 < title${title}.sub
  done

• Convert every subtitle to image

  #! /bin/sh -e
  for title in `seq 1 6`; do
  	mkdir -p itcrowd${title}
  	vobsub2pgm -t 2 itcrowd${title} itcrowd${title}/itcrowd${title}_
  	for image in `find itcrowd${title}/*.pgm`; do
  		convert ${image} ${image}.pcx && rm ${image}
  	done
  	mmv "itcrowd${title}/*.pgm.pcx" "itcrowd${title}/#1.pcx"
  done
  

Decode

The subtitles from episode 1, 2, 3 and 5 contain base64 encoded files. It's possible to use tools like base64 from coreutils - but they can only handle text files. So get 100 secretaries and let them convert each file into a text file. ;-)

I "created" a small program which memorize every letter I converted manually and does it the same way I did when it finds "the picture" again. So I had two runs - one to fill the brain of this little dump OCR software and another to save the result.

If you want to start with an empty brain.db (no, you don't want that... trust me), don't pipe the output anywhere because the first run is an interactive run (full brain.db is inside the archive). Afterwards it is intelligent enough and you can convert everything in a separate file.

#! /bin/sh -e
BRAIN=~/brain.db
for title in `seq 1 5`; do
	for image in `find itcrowd${title}/*.pcx`; do
		itcrowd_decoder ${BRAIN} ${image} > ${image}.txt
	done
done

Merging

It's important to know that some subtitles are duplicates and must be removed before we can continue by merging everything. There are different strategies to do that... for example by looking at the timestamps. It's a possible duplicate when there is no difference between the end of the last subtitle and the start of the new one. I used a small script to do that.

#! /usr/bin/env python
dict = {}
lines = open("itcrowd6_.srtx").readlines()

i = 0
cur = ""
last_end_time = ""

for line in lines:
	if i % 4 == 0:
		cur = line
	elif i % 4 == 1:
		t = line.split()
		if t[0] == last_end_time:
			print cur

		last_end_time = t[2]

	i += 1

• itcrowd1_*.pcx.txt
  • 0125
  • 0140
  • 0236
  • 0433
  • 0494
  • 0538
• itcrowd2_*.pcx.txt
  • 0127
  • 0363
• itcrowd3_*.pcx.txt
  • 0009
• itcrowd5_*.pcx.txt
  • 0002
  • 0187
  • 0378 (needs to be replaced)
  • 0751
  • 0902
• itcrowd6_*.pcx.txt (different tool to decode that into txt files)
  • 0002
  • 0298 (needs to be replaced)
  • 0602
  • 0877
  • 1366

After we removed the duplicates and cat'ed the remaining ones into itcrowd${title}.txt, "base64 -d" can do his work (don't forget to remove MIME-Header from the beginning of the file).

EasterEgg1.jpg

• picture of Moss
• Exif comment: You've found an Easter Egg!
• Exif GPS-Position: Pinewood Studios (I.T. Department set)

Barber.z5

The second episode will bring us a game for z-machine interpreters like frotz. My personal walkthrough to get 42/42 points (there are some bugs which makes it even harder):

                  .--------.
                  |        |
                  | Toilet |
                  |        |
.------.----------o----#---o---------------------------------------------.
|      |          | Toilet |                                             |
| Lift # Corridor #        #              End of Corridor                |
|      |          | Area   |                                             |
o------o----------o----#---o------------o---------o-----------o----#-----o--------.
                  |        |            | Storage | Moss's    |          | Jen's  |
                  | Toilet | Kitchen    #         #           # Entrance #        |
                  |        |            | Corner  | Desk      |          | Office |
                  o--------o------------o----#----o----- -----o----#-----o--------o
                           | Richmond's | Roy's   |           | Games    |
                           |            #         # Workbench #          |
                           | Room       | Desk    |           | Corner   |
                           o------------o---------o-----------o----------o

• Lift
  • take screwdriver
  • press button
  • wait
  • east
• Basement Corridor
  • east
• The Toilet Area
  • east
• End of Corridor
  • take spectrum
  • check keypad
  • read note
  • type 3584 on keypad
  • south
• Office Entrance
  • take zx80
  • east
• Jen's Office
  • take computer
  • switch off computer
  • switch on computer
  • west
• Office Entrance
  • south
• Games Corner
  • take xbox
  • west
• Workbench
  • take radio
  • north
• Moss's Desk
  • take commodore amiga
  • west
• Storage Corner
  • take bitstik
  • west
• Kitchen
  • east
• Storage Corner
  • west
• Kitchen
  • take pendant
  • check ceiling
  • east
• Storage Corner
  • south
• Roy's Desk
  • west
  • say I am your boss to roy
  • open red door
  • plugh
• Richmond's Room
  • check lights
  • east
• Roy's Desk
  • use computer
  • plugh
• Richmond's Room
  • check lights
  • check lights
  • check lights
  • check lights
  • check lights
  • open red door
  • east
• Roy's Desk
  • east
• Workbench
  • east
• Games Corner
  • north
• Office Entrance
  • open door
  • north
• End of the Corridor
  • west
• The Toilet Area
  • type 5982 on keypad
  • north
• Toilet
  • take two-way radio
  • switch on two-way radio
  • tune two-way radio to 159.65
  • listen to radio

EggDump

• bbc master mode0 image
• can be decoded by using SOPHIE or into a PPM by

  #! /usr/bin/env python
  data = open("EggDump").read()
  
  print "P1"
  print "640 512"
  
  def print_byte(x):
  	for i in range(0, 8):
  		print ((x & 0x80) >> 7)
  		x = x << 1
  
  def decode_blockline(x, y, line):
  	byte_pos = y * 640 + x * 8 + line
  	byte = ord(data[byte_pos])
  	print_byte(~byte)
  
  for y in range(0, 32):
  	for line in range(0, 16):
  		for x in range(0, 80):
  			decode_blockline(x, y, line / 2)

SOPHIE

A small bbc basic program - so we need an emulator for bbc machines. I tried BeepEm to simulate the basic interpreter and the acorn hardware. You can load the image with the basic program and EggDump from the previous episode, type CHAIN "SOPHIE" and press ENTER to watch a picture of Roy with an easter egg and hear the it crowd melody.

DFS Explorer can be used to create that image (I used the Linux freeware version). It is extreme important to set Load of EGGDUMP to FF3000 and Exec to FF7FFF and save SOPHIE as basic program (don't know how to do that without typing everything into the basic shell and saying SAVE "SOPHIE").

More about the usage of the basic shell can be found in the official User Guide.

EasterEgg3.part.2

After watching the episode again with mplayer, I noticed that subtitle page 378 differs from the one I got from subtitleripper. Even with SubRip it wasn't possible (ok, this thing exported nothing as it should) to rip the subtitle image. Only a typing monkey could create a working version which filled the missing bytes.

You cannot use the data before you didn't extract EasterEgg3.part.1 from title 6.

Thornes box

This one is very different. No base64 encoded stuff all over the screen - only a small box in the lower right corner. If you have watched the (hidden) track 10 on the DVD, you will (maybe) know what that means. I tried to think a little bit about what each line could mean

9 bits is a little bit too much - so maybe the lower bit is a parity bit (i tried to check it by decoding everything and it looks quite good). It seems to work the same way the overscan does (with an extra empty line after begin and before end marker). So mplayer -vf crop=1:42:719:38 ../itcrowd6.vob -vo png -nosound can be used to get the "raw" overscan images.

And again, I miss two bytes. Where are they gone? You can look in the subtitle blob at offset 1002 and see a nice message and guess where one byte is missing (0x49 at the beginning of the file). The other missing byte ruins the image file directory at offset 1186 - so the missing byte must be in front of it. After some research, I noticed that subtitle2vobsub or vobsub2pgm is buggy and we can replace itcrowd6_0298.pcx and some other broken exports with a version I saw by using mplayer. This one will give us another hint what to do next.

The overscan has many blank frames and some extra frames to learn the encoding (frame 36-290) which must be removed. After the conversation you will get EasterEgg3.part.1 with some greetings at the beginning of the file and a checksum + file length... wait a second, I am missing a byte again?!?

This time the frame for offset 17708 is missing. Just insert 0x03 and everything looks fine.

You can cat both parts (without the comment at the beginning) together and get a picture of Jen with her easter egg. Take a look at it with a tool like exif to get more information out of the image - like the image unique id. Looks suspicious, or? This hex string can easily converted in a ASCII string (use an hex editor if you don't know all ASCII chars hex representation). If you done it right it should give you a smart hint for the competition - 'A: "Theobromine"'.